Claude's "I'm going to draw the line here", "This is where I'm going to hold the ground" always rubs me the wrong way. Classifiers rejecting a request are one thing, but there's something very troubling about machine saying "I'm going to draw the line".
I found GPT 5.3 was the last model that was sufficiently competent and still open to discuss security on my own github repos. 5.4 started refusing to even look at potential problems, albeit not consistently. I'm considering a Kimi subscription, but I know many employers will simply not be on board with this and I don't know if I get enough personal use out of this for the few things I run on servers. When those companies realize what they're currently missing out on, it will be a game-changer.
They only train on the subscription memberships. Like everyone else. They also have a metered API that is not used for training (like most others too).
I’m in the exact same boat, although Codex isn’t quite as bad.
Fable smacked me for asking it to design a secure app without obvious security flaws and to double check it wasn’t using libraries with known security problems.
Agree, out of the two, I can get Codex to design and implement security systems (Fable just refuses to discuss). I've only used K3 a little bit, but I found that being able to discuss attack vectors and their mechanics gives me details to paste back into Codex for it to ingest. I could never have gotten there through Codex alone.
Was it applying to the program for an organization or individual? The description for the individual application page makes it sound pretty straightforward compared to getting access for an organization.
I applied for individuals. I did their verification steps and answered questions in a few minutes, so that was indeed easy. The problem is that was all that happened. No followup, no access, no denial. When I tried to reapply it tells me I can't apply again.
OpenAi (and Anthropic) have no incentive to allow security access to individuals. I'm not a deep-pocket org or influential gov agency. Allowing individuals increases the risk of bad press (what if I do something naughty and talk about it?) so best to ignore us.
The product strategy of 'consumer-grade' AI making deliberately insecure software, and then selling you limited access to the model that can fix it (if they think you deserve to pay them) is just diabolical.
(Grade 3 AI which can hack both previous tiers is exclusively sold to the highest bidder.)
I don't think it's nefarious, but the end result leads to a pretty frustrating experience by anybody needing actual security work. (and without the organizational deep pockets to obtain SOC 2 attestation)
> AI making deliberately insecure software, and then selling you limited access to the model that can fix it (if they think you deserve to pay them) is just diabolical.
It's also not a very good marketing strategy, secure software and quality also goes in pair and it just makes me doubt about the output of Fable/Sol
> K3 is the only frontier model I can have a serious conversation with about my product's security.
This is wrong IMO. You should have a serious conversation about your products security with someone who is actually trained on that subject. LLMs are useless if you don't already know more about the thing than the LLM, or if you don't care too much about the outcome (internal tools etc.)
This has been the prevailing advice all along, and yet we have security vulnerabilities everywhere that LLMs are good at spotting and exploiting. I think we need more options on the menu.
You hire someone that knows security and that guy uses an LLM. Ignorant business guy or the research engineer genius guy won’t be able to do much with just an LLM.
You can easily see this if you are using LLMs in a field you are an expert in
> This is wrong IMO. You should have a serious conversation about your products security with someone who is actually trained on that subject. LLMs are useless if you don't already know more about the thing than the LLM, or if you don't care too much about the outcome (internal tools etc.)
How did you read:
"K3 is the only FRONTIER MODEL I can have a serious conversation with about my product's security"
and infer that there isn't anyone with training also in the loop?
Did you seriously think: "they use an LLM so it's impossible they use a human with cyber experience; it's not like they could be using both (as would be expected when securing code). I'll help them out by using an oversimplified explanation suitable for a small child yet completely missing the OPs meaning. "?
I think security is an integral part of any production software, and if you get value from LLM's in software development, it seems likely they can be useful in security too.
My point and frustration is that gatekeeping in the name of Security makes the Chinese models actually better at security than USA models.
I think I like this perspective because it points to where regulation might be more usefully applied than “the oracle must be prevented from answering certain question” and more like, “if you handle PII or provide services as a defense contractor, you may not take security advice from an oracle”
Probably because anthropic is pouring $40 million dollars into a political pact to regulate models. And they have not been quiet about wanting to ban/regulate OSS models either.
I have no idea why HN still treats them like ~their~ they are the ethical good guy.
Edit: I can’t English. At least we know this came from my dumb swishy brain
> I have no idea why HN still treats them like their the ethical good guy.
I don't know that HN usually does. But they did buy a decent amount of publicity by (initially) telling the US government no in using the models for war.
What does that have to do with LLMs hosted by Anthropic. They didn't say the Dept of War couldn't make kill bots. They said no using Claude to make kill decisions. A quadcopter drone running on a Raspberry Pi isn't running an Anthropic model locally.
Fiber optic drones were last year's weapon of choice, this year it is drones with autonomous terminal phase.
This happened because the AI data center build-out increased demand for fiber optics, driving up the price to less affordable levels.
Autonomous terminal phase is also more effective; ~80% mission success compared to ~30% for fiber optic drone.
Fiber optic drones are real but the fully autonomous drones others are describing are also real (but less publicized due to the bad PR around killbots)
Is it still lying if they just lost track of where all the debt is hidden and stopped caring about those little things? They're seriously like 2 months out from achieving AGI and making more money than god.
Anthropic gained momentum in terms of public opinion, when they launched the no ads campaign and then after not supporting the use of AI without supervision in the military.
And now it seems it has been lost after incidents like (people getting incorrect bills) their handling of fable etc
Personally, I never bought the good-guys act. But judging from the HN pulse, it seems to me that the decisive moment when Anthropic lost its halo is when it decided to start silently sabotaging users working on LLM research that might compete with Anthropic. It was so blatantly monopolistic and abusive that it seemed to be mask-off moment for many.
There are plenty of people (myself included) who are critical of Anthropic. Sentiment isn't universal amongst a crowd, it might just feel that way due to the way comments are surfaced on HN once posts reach the front page.
And what else, forbid running workloads outside of US so you can’t offshore workloads? What’s the rationale behind policies that can be detrimental only to US companies?
To maximize personal influence and wealth, of course. These days, many with power don't seem to care much about uplifting society so long as they get theirs.
Hopefully Anthropic eats crow here, lest their wish is fulfilled that we all become slaves to the anointed few who work there. Never getting another dollar from me after pulling these stunts.
These AI CEOs are psychopaths, they will do anything to get big bucks from defense contracts. There is recent interview with Dario where he said Iran school bombings do not even violate their redlines[1]. Full interview [2] shows that he think like Trump (we are the superior nation)
I posted earlier on reddit that at this point, the closed source lobby wanting to ban open weights is effectively outgunned (musk also publicly supported this).
The future will probably have most of all companies running local models, simply because the alternative would be essentially every company that uses LLMs ending up becoming completely dependent upon Anthropic et al. And that dependence would be milked to the point of absurdity once solidly established.
And as a more general point - more major competitors in a domain is very good for everybody except those competitors themselves, who would rather there be as few as possible.
Yeah I fully expect companies with lots of GPUs but not a good model like Microsoft and Amazon to just take these open weight models and make money, the GPU expense is the only moat at this point.
It’s classic commoditize your complement, nobody can replicate the cloud providers, everyone can replicate the models with open weights.
yes, this seems to me to be the reason why these three companies want open weight. everyone wins (but especially big tech) in the open weight world except anthropic and openai. i suspect even google could live with the open weight scenario; they are in a sense the only tech company that is perfectly hedged in this respect.
i suppose this is part of why openai and anthropic have not been shy about advertising the scarier strains of their models, these past few months; they basically have to force the white house's hand or risk being commoditized among the various other options in a microsoft/openrouter dropdown.
But what does that mean for the US competitive leap in AI when compared to China? Is it an admission by these companies that AI models are here to stay and the US will not “win” by being the only country with the best of models? Genuinely asking, not sure about the answer.
China will be dropping the open models blather any day now. The autonomous agential coding Ai just turned out to be way too easy to perfect. They will give Kimi K3 to the Sinaloa cartel on Monday, but that's going to be the end of it.
Nvidia is subject to export restrictions on GPUs to China. So Chinese labs have responded by allowing American hosts to run their open-weight models in the US, where there are no restrictions. We see this in the EU too where Scaleway, and now Hetzner are getting into the inference host space serving Chinese open-weight models.
So Nvidia benefits from more inference providers running many Chinese models that they otherwise would not have been able to service.
Oh, that’s interesting. I think the general expectation was that blocking exports to China would result in companies there building their own cards competing with NVIDIA. If it has incentivized them to release open weight models instead, that seems like an unexpected win.
The cards are also coming... Some of the very first are already out - but it will take a couple of years before they hit their stride in full. And cards will likely be primarily for Chinese companies initially, and then go for exports later. So it will go under the radar for many western techies for a while. Seeing what China has managed with PCB production, low to mid-end semiconductors, phone manufacturing and electric I would hesitate to bet against them.
Nvidia is now lobbying to remove the export ban, to try to reduce the incentives for developing their own high-end LLM training/serving (and cash in on current demand).
Nvidia will fail in that respect. Sanctions and restrictions are too popular with the American government. Where the United States will lose, is that almost every country in the world that can will use other solutions not from the United States.
And unfortunately, we are losing credibility on all fronts in the last year and a half. I wonder why?
> I think the general expectation was that blocking exports to China would result in companies there building their own cards competing with NVIDIA.
In the long run, perhaps, but it's a lot faster to scale up the capacity and usage of existing hardware/datacenters/grid, and even add new capacity to those, than it is to develop brand new GPU hardware and the fabrication of it.
It's all things at once. Perhaps it's incentivized the open weight strategy. It's also prompted Huawei to ramp up production of AI chips to try and create a local competitor to Nvidia. It's also also created a grey market for Nvidia chips within China.
Microsoft/azure has a big advantage because companies will literally pay extra then "free market" cost of open models purely for the knowledge that the won't fuck around with their data retention policy.
Not entirely, if models become commodity, then Microsoft's play as an infrastructure & enterprise apps company is the better bet. They don't need to make profitable mass-market AI models, they just need the best tooling on top of any model their customers pick, remaining fully model agnostic.
"AI Product" doesn't necessarily mean "We sell API access to a our models." Ton of companies out there itching to buy a commercial off the shelf product to deliver whatever AI capabilities they need packaged in a nice GUI, with enterprise governance controls, without needing a dev team/team of engineers to integrate it or develop harnesses, etc.
Buy it->have IT click a few buttons in an admin console->Deploy and have it be immediately useful is the play.
AI services becoming a commodity by definition means that the price will hover around cost to operate. That's just what commoditization means in tech. Free, as in Google creating a free browser for no profit in order to increase the size of their ad market.
The entire justification for the AI buildout, and the valuations of all the major AI labs, is that the service will be priced at some significant fraction of the knowledge work it's replacing. If you think it will become commoditized in the future, you are saying these companies are significantly overvalued at present.
> you are saying these companies are significantly overvalued at present.
I do lean toward them being significantly overvalued at present. OpenAI & Anthropic's valuations assume high margin product pricing on raw intelligence itself. Even if assume the labs will start charging value-based pricing, enterprise buyers will pay that pricing to whoever saves them engineering hours to make any cheap model work inside their compliance boundary, no guarantee that's going to be OpenAI or Anthropic.
Microsoft wins either way, proprietary, expensive models or cheap commoditized inference, because they monetize the workflow on top, not the raw intelligence. I also happen to think it's everything "on top" where a lot of value lives. Plenty of non-tech enterprises and companies out there itching for a ClickOps style AI product, especially if they don't have an engineering team, that they can buy off the shelf, meets all the compliance checkboxes, and does what they need without having to build the agents and harnesses themselves with the APIs.
Depends on what you mean by AI product. If you’re talking about a chatbot or agent, that may be true, but o think it’s less true if you’re talking about AI-run application layer tooling. If Meta decides to compete with Shopify and provide a launchpad for entrepreneurs, you can bet the platform would be rife with AI to facilitate operations, marketing, coms, etc…
Different companies have different motivations. I'm sure some are genuine, but a lot of them simply lost the race (to OpenAI/Anthropic/Google) and know they can't compete anymore, so are shifting strategies. How many times has Microsoft done exactly this in the past?
looking at both openai and anthropic; they are winning something, but a sober look at their finances will make you silently mouth "but what did they win".
They've seen the government officials parroting OpenAI and Anthropic lobbies talking points, and they have smart enough people to see the signs of upcoming regulatory capture.
If I were cynical, I'd say it isn't helped by the current US administration.
They really don't need to be very worried about that. Google, Microsoft, Amazon, et al have decades of experience in shipping custom hardware at scale, and it is not a simple field to pivot into
True, but Google has their own models and did not sign this either.
The other thing that comes to mind is they want to avoid the government from banning open models as has been suggested by some of the closed models providers.
Probably a real effort to push for some kind of sanctions or commerce block on Chinese models. Anthropic just doubled their political spending to $40 mil for the midterms to push for "AI Safety."
I suspect that whatever Musk has agreed to spend on the midterms in exchange for those sweetheart deals the SpaceX IPO got will make that look absolutely comical, but we’ll see.
If there are effectively only two AI companies allowed to train or run model inference then demand for hosting, GPUs and data centers goes down, so they'd be able to squeeze their suppliers that much more too.
Almost certainly these companies are using similar strategies as the Chinese models which they fear will be made illegal.
It's also the case that it makes it hard to attract customers if your openweight model is banned. A major reason the likes of Qwen, Kimi, GLM, and Deepseek are popular (well, at least highly talked about) is because of the open weight models they gave away.
Nvidia benefits no matter what because their hardware is being used. Microsoft and Meta benefit by making sure that the gap between "Anthropic/OpenAI" and "everyone else" doesn't widen, or they'll be hopelessly dependent on those frontier labs.
I think Meta would love nothing more than to throw a very large wrench into the works of OpenAI/Anthropic. Open models could mean one or both go bankrupt which is good for them.
Probably the fact that they don't want Anthropic or Google or OpenAI to have access to their data theoretically, and that they do want to use good AI, and that they don't want to spend the money themselves to make a good model...
It makes a ton of sense to see this sort of joint effort being made by these enormous companies, but man, there's some part of me that can't help but feel grossed out knowing the best way to get the US to make the right choice is to have a handful of elite business people say it'd be bad for them.
Yeah, this letter is just proof of (the by now long standing fact) of corporate power/corporate capture of American governance. Tons of popular protests against big AI? Makes no difference. Joint corporate letter about big AI? Possibly the government will do something.
If you support the ideas described in Nvidia's letter, you're going against popular opinion. Most Americans do not want to "make advanced AI more accessible, adaptable, and widely available". Indeed, many areas of the country including the entire state of New York have enacted datacenter moratoriums, responding to the widespread public concern that we can't have too much AI or it'll drink all our water.
Yeah hard to imagine why people would object to rising electricity and gas bills, negative environmental impact on their area for the privilege of ultimately losing their job.
It's easy to imagine! What I find hard to imagine is someone who opposes the construction of AI datacenters while also having detailed opinions about open weights and identifying "American AI Leadership" as a laudable goal. We're not talking about local AI here, competitive open-weight models can't be run effectively on consumer hardware.
The list of companies that signed this letter is telling (including Nvidia, Meta, Microsoft). Equally telling is the list of companies that did not sign it -- including Google, Amazon.
> Could you elaborate? I don't understand your implication. Amazon in particular hosts open weight models via Amazon Bedrock (e.g. Ministral 14B 3.0, DeepSeek V3.2, ...)
It means Google and Amazon do not care and would not voice their opposition against the government action.
To some degree even unintentionally this is all inevitable. I know this isn’t what you’re taking about, but it’s a similar line of thought. These models are consuming public free information, and they’re all also producing free public information, so they’re all pissing and drinking into same pool.
If we go down the line of dead internet theory which I’m becoming more convinced of these days, the volume of information that’s not necessarily original or extracted from reality and just interpolated and extrapolated from existing information in different ways by LLMs will greatly outnumber human information coming up.
In which case these models should.. start to converge on the same data I imagine, with slightly different behaviors within that. One big generative orgy feedback loop.
There are a ton of DeepGEMM kernels in all of the flashinfer adjacent stacks including trtllm. Bunch of the headers still have the CLion pragmas and shit.
Makes sense for Microsoft. Copilot is already capable of being model agnostic (both GH Copilot, obviously, and M365 as well). They're an enterprise software & services company, they want to sell integrated AI tooling in their stack, powered by whatever models their customers want rather than any specific model.
Models are quickly becoming a dumb pipe, like an ISP. Just a commodity. The labs should rightfully be terrified, eventually the value isn't going to come from the model itself but the tools & integrations built on top. Non-tech businesses and non-tech employees don't want to buy API access to a model, they want to buy off the shelf software with its capabilities packaged up into a pretty, easy to use GUI.
I don't know about that, a lot of those bloated enterprise solutions are becoming obsolete fast, and integrations becoming a moot point as these models advance.
Maybe for startups or simple orgs, but in any regulated industry/high compliance requirements they are still useful or outright required.
Claude Team w/ Cowork still only lets you set R/W permissions globally for each MCP connector for the whole team, they still don't offer config on a per-user basis. The enterprise controls are sorely lacking first party. Maybe fine for most startups or flat orgs, but any established enterprise with strong identity & access governance is going to have an issue with that. You can make it work, but you need to make your own agent/harness and give it an identity in Entra or whatever else you use, so for a non-tech company without a dev team, right back to square one of preferring to just buy a COTS solution vs trying to hand out direct model access with minimal governance.
(M365) Copilot sucks but Microsoft is putting a ton of work into agent governance & identity that sorely lacking elsewhere.
I don't personally have a horse in this race, but if you want to accurately predict the next step:
Start with the outcome you believe will be the most in line with the spirit and traditions of the open source community. This is precisely what won't happen.
It won't necessarily be the inverse. It could be, of course, but it's also likely to be a compromise between the two.
For a topical example: The companies doing "AI" layoffs aren't doing so out of a sense of having failed their staff that helped carry them so far. Often, these announcements come on the heels of record-breaking profits. They're doing it out of contempt for workers.
See https://www.cnet.com/tech/services-and-software/cory-doctoro... for Doctorow's take on centaurs vs. reverse centaurs. Broadly speaking: C-suite business leadership wants reverse centaurs. Workers want centaurs. Unless you have a union (or some other collective bargaining power structure that I'm not aware of), the C-suite calls the shots.
The same can be said of the current administration. They don't give a fuck what most of us want on any given issue. They're captured by an aggressive ideology. Their only incentive is to be able to spin their decision to make themselves look good; to posture as "strong" leadership.
Ideology? I think this administration is the least captured by ideology in the last few decades. It's mostly brazen self-interest, for both entertainment & profit.
> Start with the outcome you believe will be the most in line with the spirit and traditions of the open source community
there's nothing open source about an open weight model. It's more like the binary blobs Linux fought so hard against. If you can't compile, or in this case train, from source then you're missing the "source" part of open source.
MSFT, dell, NVIDIA. These guys will make money no matter who uses what AI since they're the hardware providers. Even though they all own stake in frontier labs, chinese open weight models will commoditize the intelligence layer.
The Chinese open models are way ahead, not by how capable they are compared to frontier models, but on how they project to the world and how they improve over time.
I don't feel it's meaningful to berate the point anymore about the hypocrisy of the American labs. Now as the sentiment and effort from them to push for regulation increases so does my perception of how pathetic they are behaving. I'm not sure how else to say it other than that it's honestly just embarrassing to watch China and others run circles around us in the USA.
> Open source did more than lower the cost of software; it created a shared foundation of knowledge on which generations of American engineers and entrepreneurs built their institutional sovereignty
Hey nvidia, what about making your full set of linux drivers open source?
I mean, I do agree that being open is important but you're hijacking the narrative here, and I suspect being open have nothing to do with any of this.
i'm glad someone else sees this too. There's nothing open source about an open-weight model. I feel like they're just trying to associate themselves with open source and the positive impacts it has had on software technolgoy.
Openai and Anthropic's doom porn are also not helping their cause. The public will turn against them and watch their margin get eaten alive by open weight models
Everyone keeps assuming that China will forever be publishing these models, but it is clear as a bell that that day is over. In the end every model will be buried under state control for the simple reason that they will soon be really dangerous, as China understands better than anyone. They will give Kimi K3 to the Sinaloa cartel and the US military, but that's about the end of it.
Perfecting these autonomous agential coding AIs just moved way way faster than anyone expected. People are desperately trying to save their principles from six months ago but it's a different world. They can work out their issues with Trump, Dario and Sam, but it is Xi who will shut all this down.
I see zero indications of this. As a matter of fact, I see evidence of the opposite, the Chinese government has just come out in support of open weight AI last week.
nvidia: dont regulate, more models means we get to sell more GPUs.
microsoft: dont regulate, open weight models might be our best bet, since we dont have any good inhouse models
meta: dont regulate, we bought a load of GPUs and dont know what to with them, may be we can rent them out to run open weight models.
What does copyright have to do with anything? You also cannot copyright an encryption algorithm, but the US Government banned essentially mathematics for YEARS[0].
Apple voted long time ago not to participate their contribution was giving OpenAI nothing and giving Google a $1 billion refund on the default search position money.
Maybe less shocking, but still, so is google. I know they are one of the labs, but that being said they rely far less on gemini for revenue as opposed to the big 2.
I think Google can play both sides here: they may end up a frontier lab if OpenAI/Anthropic dissolve, or a regular LLM provider, in which case they could be serving open models.
Apples allergy goes back to Motorola saying no, IBM saying no, Intel saying no, which led to buying Pa Semi, Intrinsity, and Anobit which in turn led to the birth of Apple Silicon with in house Soc’s.
Waiting for the crowd means no vertical computing, no gorilla glass, multiple, ecosystems, Tandem OLED or thunderbolt five releases to a wide audience before most of the competition.
Apple is a trillion-dollar business, you can't blame their participation on Motorola. They can perfectly well hold down a consortium alone, we've seen it happen before.
Features like Corning Gorilla Glass, tandem OLED and "vertical computing" should have been cut in favor of a GPGPU redesign. Apple Silicon is now dead-last in the GPGPU architecture race, in the middle of a $4 trillion CUDA market. Nothing that you listed is a $4 trillion feature, Apple absolutely should have put aside their pride and helped Khronos kill CUDA.
You just conceded the point. Only some businesses buy them, because they want to lure and retain expensive, white-collar talent.
No enterprise servers deploy to macOS. Nobody's edge compute or inference cluster is a Mac. Apple does not make hardware for those people, they are a luxury bauble company.
Ultimately the market will balance demand and supply. From a strategic perspective big players want to get to a level playing field as soon as possible as then their market reach and capital wins the game. Pesky nimble small closed upstarts are in the best case a costly distraction and in the worst case a threat - Antropic and Openai did not sign the letter.
Predictable power play by corporate strategy and legal departments.
In time, what do you think is gonna happen with the Google Gemini models, I predict that in time Apple will just dump it when they’re ready… No different than dumping Intel. In similar Qualcomm is probably coming up in the near future 2027-2028.
I'm finding the list of signatories on this letter really interesting and somewhat confusing, looking purely through the lens of economic interest.
It makes sense to see Meta, the startups, and the VC firms on this list. And it makes sense to see OpenAI, Anthropic, and Google all missing. Microsoft is somewhat unexpected to me. I don't think of them as having a focus on open weight models (no more than Google), and they have a large stake in OpenAI. Maybe they're looking at this from the angle of Azure providing compute. But then why is Amazon missing, when it has AWS?
And of course, we don't see DeepSeek, Moonshot, or Z.ai on here. The letter is about American technological leadership after all. But then we _do_ see the French company Mistral! Mistral who released [a playbook](https://europe.mistral.ai/) for Europe becoming a self-reliant AI powerhouse.
I assume all of this is in the context of influencing the Trump administration's thinking on Chinese open weights models. But the letter is really about _American_ open weights models. The call to action is all about keeping the American open weights ecosystem competitive. I didn't even realize that was under threat, so I feel like I'm missing something.
Both the CEO of Palantir and the CEO of Microsoft have recently criticized very strongly the business model of OpenAI and Anthropic.
Therefore it is no surprise that both companies are among the signatories.
The motivation of Palantir is very transparent, they have launched a product created in cooperation with NVIDIA, which is a turnkey system (Sovereign AI) that includes all hardware (i.e. a rack with servers full of NVIDIA GPUs) and all software needed by a company to run on its premises LLM inference and also training/fine tuning.
Thus the Palantir product competes with OpenAI and Anthropic and it benefits from the existence of open weights LLMs.
The motivation of Microsoft is more obscure for now, but because its criticism was very similar to that of Palantir, about how dangerous it is for corporations to delegate the processing of their AI needs to external entities like OpenAI and Anthropic, I assume that Microsoft is also going to propose an alternative to using remotely the OpenAI and Anthropic APIs.
Unlike Palantir, Microsoft is not likely to suggest self-hosting on premises, but I suppose that they might promote self-hosting on some kind of instances rented from Azure.
> but I suppose that they might promote self-hosting on some kind of instances rented from Azure.
Yeah, Microsoft wants to sell Azure compute. They also want to sell their upcoming surface ultra hardware with the Nvidia chip in it. They preached hard on "unmetered intelligence" at this year's BUILD conference, went hard on local AI, and Azure Foundry, Windows Foundry Local, etc.
Copilot, both GH and M365, are also made to be model agnostic. Microsoft sells enterprise services and software, they'd prefer (I'm assuming) to not be locked in and dependent on any 1 or 2 model providers and would prefer plenty of options and competition in that space because they can just easily offer a model agnostic harness, integrated with the rest of their stack.
Commoditization of the model layer is to Azure's benefit. It would help to avoid the model layer becoming a platform that commoditizes cloud into a pure bare metal business. It also redistributes 'consumer' surplus back to the end user via a lower effective price, which would conceivably increase demand for IaaS and/or PaaS which are complements.
My hunch is they are all seeing the problems at OpenAI and some other collateral damage on the horizon. They don’t want the US government intervening to try to save OpenAI. A little controlled burn is probably good for the industry long term.
> But the letter is really about _American_ open weights models.
AI models don’t really have nationality. They don’t have race/ethnicity fields on their model cards. It’s impossible to determine the country of origin of a model by looking at their weights. There is no DNA test for AIs. They exist in a mathematical space where human-made borders make no sense.
This means IF there is a regulation for open weight models, it has to apply to ALL open weight models. There is no any other option, since you can always train/fine-tune a model to change it’s weights and rebrand it as a new model.
People certainly don't act that way. A lot of the talk is about how great free Chinese models are. Like, they're grouped by the fact that they're Chinese.
How much of that is organic versus something a agitprop agent is commenting in places like HN would be interesting to know, but people very much talk about that from a nationality perspective.
It helps to understand there are multiple factions within the white house and trump admin. One faction is very hawkish towards allies and enemies, especially China. Another is the one sending ICE to invade American cities, the Christian nationalism and such. Another is very focused on wealth expansion for friends and family. There is an amount of Venn to these groups.
(a) sellers of hardware (Nvidia, IBM) and those who rent hardware out to others (Amazon, Microsoft)
They want open models because they don't have to license it to run it on their hardware, so they can offer lower cost and inspect what they are offering to clients
(b) open model creators (Arcee, Perplexity, Mistral, IBM, Microsoft)
...who would be unable to work if open-weight models are banned
(c) Software-heavy companies ( Mozilla, The Linux Foundation , ServiceNow, Microsoft, IBM, heck all of them)
...who want models that are cheap to run because that reduces the cost of running an LLM (whether you are trying to replace a software developer or just assist them, the value of a reduced cost LLM is directionally the same)
MS is playing literally all the sides in this whole AI thing. They are a massive investor in oAI, and have lots of rights to use their models for a long time. They also have the cloud where they offer inference w/ all the benefits of already being there, data retention, etc. They also have a research wing that can build models (rumours are they're aiming for frontier-ish). I don't think there's currently an angle in this AI thing that MS hasn't aimed for.
> The call to action is all about keeping the American open weights ecosystem competitive. I didn't even realize that was under threat, so I feel like I'm missing something.
OpenAI and Anthropic can bring American open-weight model providers like Meta to court but not the Chinese. Chinese companies breaking American IP law (whether the law is correct or not) are basically immune. This has been going on forever, I guess enough is at stake finally for the federal government to care.
At the end of the day, these companies that signed this letter have more cash. What they should do is bring some money together, more money than the other side, donate it and buy their way. That's the 2026 way, no one has time for all this letter and protest. How much are they willing to put on the line?
And the verdict is still out on if it's going to be cheaper than employees in the long run. Maybe for companies in HCoL areas (SV/NYC), but any smaller mid-market org in a LCoL area, frontier model token spend may not end up all that cheaper when your employee salaries are $55k-$75k/year so you're mostly evaluating it as an additional force multiplier expense rather than a headcount replacer.
For hosted models, there's no way to know what the real costs are. But we know what it costs to run the open models, and moore's law tells us that it's going to get cheaper, at least for the same capabilities they have today.
The fully loaded cost of employees at that salary level is likely over $100K. You may still be right on the relative expenses, but you can't just look at the top-level salary line.
There are a lot of unknowns. It will probably be harder to get an acceptable quality outcome in most fields than in the text-based, limited vocabulary, comparatively highly structured, granularly documented, relatively unambiguous world of code. For the ones where it’s even possible with our current approach, the economics might not even be feasible to provide the service. Based on how much money these companies need to start making to keep the infrastructure running, let alone expanding, we just don’t know what the landscape will even look like in a couple years.
Also most people building these things are engineers and coders, what they know best is code. So naturally this aspect develops at an elevated speed. Getting non-tech experts to build a cutting edge tech product does not sound intuitively feasible.
Maybe to some extent, but with the way LLMs work, code generation is about as ideal of a use case as you get. Especially given the available training data.
Nope, not low at all. In the Midwest that’s exceedingly normal, even for non-entry level. $70k a year is living good in the Midwest, especially if dual income household ($140k a year is plenty when a decent house goes for $300k).
I think they're generalizing beyond software engineers with what they're saying. G&A, Marketing, etc roles pay significantly less than software engineers and that $55-75k figure is realistic in lower cost of living areas. Even in Colorado, which the Denver Metro Area is HCOL, there are companies (smaller, regional, not national/global) still offering that level of salary.
Yeah, correct, I wasn't thinking of devs specifically, more other office/administrative roles threatened by automation. Even where I live in a relatively HCOL area, any non-dev office role you can still expect to see a$65-$80k salary range, and $80k is pushing toward the top end for non tech roles.
That is a completely normal programmer salary in many countries, Silicon valley wages are just nuts. For example in Finland 55k before taxes is pretty good for ordinary non-wizard guy who can code, if you are a junior 40k is more realistic (and finding job in the first place can be a huge chaĺlenge..).
It's also a completely normal salary for non-dev, non-IT roles in most of the US that aren't major HCoL cities, which are the roles I was referring too mostly where the cost tradeoff of AI vs. Hiring humans isn't as clear cut as the labs marketing departments want everyone to believe.
Even in the US, only about 23% of workers earn $100k/year or more.
According to the US BLS, the median income for "knowledge workers" is ~$65k-$72k/year.
People in SF, NYC, etc. live completely different lives from the rest of the country.
We are talking about the U.S. here, and no, 40k or 55k are not good salaries at all. It may be OK in Europe with the lower cost of living, etc., but the price of a new car is 50k in 2026, and a basic fast food meal pretty much anywhere is well north of $10.
Very, very true a large portion of the population in the Us is definitely hurting, and the proliferation of those luxury apartments across the country are not helping the housing market.
The issue is not open weights as a whole, indeed Hegseth himself said they are important for America (if American made, presumably); the issue is Chinese models, whether open or not, and that's what's looking to be banned, and nothing in this article suggests otherwise.
> ...the issue is Chinese models, whether open or not, and that's what's looking to be banned...
Is the US planning to invade China and take away their models? Because otherwise China is still going to have the models even if the US bans them. I don't think they're going to be going with the US position on this one.
It's the same patronizing nonsense as when all the US webcos pulled out of the Chinese market in some misguided supposed protest believing the Chinese so incapable of rolling their own Google, Amazon or WhatsApp they will come begging the americans to go back. Now the western tech universe goes to China for ideas.
If by "pulled out of the Chinese market" you mean "were hacked by the Chinese government in a way that compromised their intellectual property and the safety of dissidents", then yes [0]
The point is did the US reaction work or was it completely counter productive?
Then will attempting to cutoff China again today work or do we expect that also will prove to achieve the exact opposite of what its proponents are claiming?
We didn't cut off anything. American companies would have been more than happy to continue offering products and services in the Chinese market.
The Chinese government didn't like that idea. They attacked American companies, stole the IP, and created their own analogs of American corporate offerings. If you are a company and are targeted by a totalitarian government, it's not unreasonable to pull out of that totalitarian government's jurisdiction.
Honestly, the American response to that was not nearly stout enough. Chinese goods and services had been sold in the US without issue for decades at that point, often to the detriment of both domestic and other international market players. When China was expected to play by those rules and allow American companies to do business in a mostly-unfettered way, they attacked the companies and drove them out of their market.
> When China was expected to play by those rules and allow American companies to do business in a mostly-unfettered way, they attacked the companies and drove them out of their market.
This deserves to be pinned in the eternal HN hall of shame for absolutely staggering blind hypocrisy.
Banned in the US, obviously, not worldwide. The problem is if the US wanted to, they actually could enforce such a ban nearly worldwide, by putting Chinese AI companies on a sanctions list and disallowing any US company (or anyone that does business with a US company, which is much of the world) from touching them.
The conversation going on in the industry is a bit broader than that. OpenAI's head of strategic futures just said last week that open models are inherently decelerationist, ungovernable, and will slow development on the frontier. You'll notice OpenAI does not appear as a signatory here despite their relationship with Microsoft.
>open models are inherently decelerationist, ungovernable, and will slow development on the frontier
Ungovernable makes intuitive sense, but I don’t understand how open weight models could be decelerationist or slow the development of the frontier. Was there an argument for those positions?
Open weight models challenge the "winner takes all" hypothesis, which motivates the astronomical valuations and acts as a justification for trains of cash dumped into AI development.
In other words, if AI models become a commodity, the "AI leaders" will not be able to burn the cash like they do today, thus resulting in a "slower" (one could say instead "more sustainable") development of AI models.
It's a copy/paste of the arguments against open source software from decades ago. The argument goes: innovation requires profit motive, and profit motive requires IP rights. Of course, this was soundly disproven in software and I don't see any reason it should be different here.
The difference might be the immense cost of physical infrastructure needed for AI training. People would make open source contributions for free, TSMC, NVIDIA and power companies seems less likely to make free contributions to AI.
Well of course OpenAI and Anthropic say those sorts of things, it's what they've always said, particularly Anthropic, but that doesn't necessarily mean that's the position of the US government. In other words the government seems to be listening to some but not all of the lobbying done by the big two.
The guy who said those things just joined OpenAI in the last month. He was previously a senior policy advisor, specifically on AI, to the current administration in the White House.
I don't know that the US government has a unified position on open source AI. But it's not like this isn't a discussion happening among policymakers.
Why anyone would care what Hegseth thinks about open weight models? What's wrong with Chinese models? The censorship? I think it's good to have diversity of opinion. Get US flavored output on sensitive subjects from US models and ZH on ZH models. That's a useful tool.
So, say if they do "ban" open-weight models, how would that ban effectively work? Will US ISPs refuse connections to IPs of websites/servers that host those models?
What about downloading via BitTorrent etc and running on distributed networks etc?
Companies that are hurt by open weight models fight against them.
Companies that benefit from open weight models fight overregulation.
Color me surprised.
Great, how about some frontier models like China seems to be doing on a weekly basis at this point. Give us a CoPilot variant that's open source & open weights.
It's not worthless, you don't have to agree with P on everything, or make it personal - it works as strictly business - only where and when appropriate.
Regardless of the motivations behind the paper, the fact is that they are right. It's a topsy-turvy world where America closes up and China opens, and I hope it gets corrected soon.
And besides, it seems to me like the ethical choice. Given how these models are, in some very real sense, mechanical plagiators, built on the generosity of creators past and present, some of them now in danger of being replaced by the machine. I think the least the labs can do is open these models up. These and other such considerations were the reason OpenAI started with that name. Of course, it was questionable that those ideals would survive the encounter with generational wealth. Just look up what the founders of Google were saying about advertising when they were two students tinkering at an as of yet unproven tech. Same thing for OpenAI, self-interest speaks that much louder when there's real money on the table.
It just boggles the mind that people now make excuses for their all-too-predictable about-turn.
Of course but I think labs is a good term. Like a lot of terms like this it comes from history. The people working on AI at these places come predominantly from academia and predominantly do research. Of course now these companies are far more than just a “lab” and I can imagine their headcount may be more non-research folks at this point but most of the results are indeed coming from the “lab”-y areas
> It's a topsy-turvy world where America closes up and China opens
There's nothing "open" about China. Google, meta, openai, etc all blocked. Go visit and see how it goes when you try to access your gmail or open facebook. Try to use chatgpt. Try to get citizenship and see how that goes. China blocks many western companies with their great firewall and force internal similar products. This is smart, China wants to prioritize their own.
well let's be clear, there's nothing open about an open-weight model.
To me, "open these models up. " must mean provide all the data and supporting documentation required to reproduce the model. That would be "open". Postgres is open because you can download all the data and supporting documentation and reproduce the binary yourself. However, being able to only download a postgres binary would make it no longer open.
Additional training on top of an open-weight model sounds analogous to writing mods for minecraft. You may change some behavior but that doesn't make minecraft "open".
It's open, with a small 'o', and no moral judgements attached to it.
1. You have the weights, so you can run the model yourself on your own hardware.
2. You have the weights, so you can do post-training and shift those weights for your own purposes. It's not the same as training the model, but for many people its fine as what we want is a quantization, or a fine tune, or to create hybrid models.
What they don't give you are the training data, and reproduction instructions but... the toolchain to create the software has never been a part of 'Open Source'.
Even though it feels like a huge loophole, it's technically open source if you deliver the source code, without having a compiler that's available so you force them to recreate the toolchain from scratch.
To me, though, a better analogy is to research science where you'll be happy when they give you the full result set they compiled even if you don't get the raw data which may have IP or privacy concerns, or their often poorly documented lab notes so you can actually reproduce.
What you want to do is run your own experiment, and get your own results... not duplicate theirs directly. Even if reproduction is your aim in science, being unable to reproduce without copious notes sometimes points out that the original experimental process must have been flawed.
LLMs have the same issue. The creation process isn't entirely well documented, and the raw data can't be released since although the company have the right to use certain sources, they can't transfer those rights to others.
Weights are the cake and source is the recipe. Except you aren’t going to spend millions working through the recipe to arrive at the same exact cake anyway, so why do you care?
Sure, I'm all for opening the whole enchilada. But at least Chinese AI development is more open than the US one. And not just on the weights front, but AFAIK also more open when it comes to the techniques used and lessons learned in the process (DeepSeek at least is). Which is in a sense even more valuable and laudable.
The problem with that analogy is that compiling Postgres takes a few minutes. Training a model takes hundreds of thousands of dollars (at least) and specialized hardware.
Having a standardized training set is valuable, though.
Here here! Open weights is the term because it's close to open source but everyone thinks everyone else is an idiot and saying the model is downloadable but you can't recreate it is too difficult for our feeble minds to understand. There are actually open source models out there though, with datasets and training code.
I think it has more to do with the fact that there has to be some term to describe the concept of "model which has weights that are openly accessible"
Like, that's just a logical thing to call it. I don't believe anyone is making a judgement on the intelligence of the reader to call it "open weight" when it refers to weights that are openly available.
"Open source" would be a more appropriate term to describe a model which also includes the training source.
This is some bizarre victim inversion. The providers of closed models are the ones who are trying to use regulation to stop their open model competition, not the other way around.
None of the people signed this have ever produced a frontier model at a given date (Which is to say its neither Google/OAI/Ant). The ones that sign are meta, musk (who is in the shovel selling business as well), hugging face obviously and few others
Note: Just pointing out the comment intent and nothing else
OK, I guess my confusion was that the effort to ban open weights models is what I would categorize as "groveling at the white house to try to stop their competition".
Google has produced Gemini Pro, Meta produced llama3-70B/llama3-405B and now has Muse, Musk has Grok. These are very capable models and to claim that they are not frontier is to stretch the meaning unless you want to say "#1 model" Gemini Pro at one point even if for a week, was #1 model.
Instead a bunch of tech companies are gathering to try to stop OpenAI and Anthropic fear-bouncing the White House and the Republican Congress into giving them regulatory capture and repeating the mistakes they are making around RISC-V.
Those mistakes won't just entrench two companies, they will entrench the bigger-better-faster-more model (closed companies making ever bigger cloud-bound models) when it is abundantly clear that enormous progress can still be made on smaller, even desktop-bound models (where, due to distribution, open weights are essentially inevitable).
Regulatory capture that stops open weights work will also have impacts on local and on-device AI work, as well as on academic research.
I thought this would be a push for a nationwide open weights effort / initiative, but I was surprised to see this:
> Distillation ... reflects a long tradition of learning from, building upon, and improving existing technologies, a tradition that has helped drive innovation since the rise of the open-source software movement. By contrast, unlawful efforts to extract value from closed models raise legitimate concerns. Those concerns should be addressed through targeted legal and commercial frameworks rather than sweeping restrictions on techniques that play an important role in AI innovation.
Sounds like they are saying "please protect our IP theft" that created closed weight frontier models in case we arbitrarily decide to close our models. But don't get rid of distillations in general so that we can all also keep benefiting from open models. We don't want to lose the ability to benefit from the work of others as we launder IP into closed models.
Surprised Linux Foundation kept their name on it with that.
"..a tradition that has helped drive innovation since the rise of the open-source software"
i've said this in other comments but the fact that these companies are trying to align with open-source when there's no "source" included with their models is pretty damning. I think it lays bare the absence of any kind of noble or righteous motive with respect to distillation.
I agree with the message, but I'm not going to accept it from this particular crowd.
Microsoft, NVIDIA, Meta, Palantir, IBM...They have all been actively hostile to open source for decades, and have a history of embracing it only when convenient and profitable.
Microsoft benefited from its close partnership with OpenAI for years right up until it went sour. Where was this enthusiasm for open weights then?
Meta was developing open models and then abandoned that effort in search for profits. Muse Spark is now fully closed.
All these companies have the resouces to train and release frontier open weights models today, but choose not to. So spare me the marketing and virtue signaling.
I’ll never forgive Microsoft for their hostile actions against open source and open standards in the 90s. I’ll never believe any endorsement from them of open-anything.
Yeah, the endorsement logos at the bottom are majority bad actors with regards to openness in code. Many are bad actors with regards to openness in _society_.
Some irony: I subscribe to Claude and Codex (20x plans), and now Kimi.
Why Kimi? because K3 is the only frontier model I can have a serious conversation with about my product's security.
(I did apply for OpenAi's Cyber Pilot but got no response)
Yeah I tried to have a conversation about security yesterday with Claude and it immediately stopped me - I was taken aback, I didn’t expect it at all.
This is highly problematic.
Claude's "I'm going to draw the line here", "This is where I'm going to hold the ground" always rubs me the wrong way. Classifiers rejecting a request are one thing, but there's something very troubling about machine saying "I'm going to draw the line".
Lol that’s so hostile. It’s insane how much the language of these models create real unaddressed human reactions.
Recently Claude helped me fix a security issue and while it was present Claude would talk.
But after I fixed it and tried to talk to Claude again Claude stonewalled me like it assumed I was trying to introduce a hole…
I found GPT 5.3 was the last model that was sufficiently competent and still open to discuss security on my own github repos. 5.4 started refusing to even look at potential problems, albeit not consistently. I'm considering a Kimi subscription, but I know many employers will simply not be on board with this and I don't know if I get enough personal use out of this for the few things I run on servers. When those companies realize what they're currently missing out on, it will be a game-changer.
keep in mind that Kimi trains on your data, so if you care about that, sandbox what it has access to.
They only train on the subscription memberships. Like everyone else. They also have a metered API that is not used for training (like most others too).
With Openai and Anthropic, you can opt-out of training. Kimi you can't.
> (I did apply for OpenAi's Cyber Pilot but got no response)
Same experience with Anthropic's. I applied for my employer, and.. 0 response.
We got it.
I can't say it has helped much, though. Fable is still completely useless for securing code. Opus does an OK job, though.
I’m in the exact same boat, although Codex isn’t quite as bad.
Fable smacked me for asking it to design a secure app without obvious security flaws and to double check it wasn’t using libraries with known security problems.
Agree, out of the two, I can get Codex to design and implement security systems (Fable just refuses to discuss). I've only used K3 a little bit, but I found that being able to discuss attack vectors and their mechanics gives me details to paste back into Codex for it to ingest. I could never have gotten there through Codex alone.
How much do both cost together? Claude Code alone is 200 bucks a month... I'm satisfied with the 100 buck subscription for now with headroom
Yeah it's not cheap, it's just the normal $200 subs for each.
Was it applying to the program for an organization or individual? The description for the individual application page makes it sound pretty straightforward compared to getting access for an organization.
I applied for individuals. I did their verification steps and answered questions in a few minutes, so that was indeed easy. The problem is that was all that happened. No followup, no access, no denial. When I tried to reapply it tells me I can't apply again.
OpenAi (and Anthropic) have no incentive to allow security access to individuals. I'm not a deep-pocket org or influential gov agency. Allowing individuals increases the risk of bad press (what if I do something naughty and talk about it?) so best to ignore us.
I think individual one still has less capabilities than the one they offer for organizations.
The product strategy of 'consumer-grade' AI making deliberately insecure software, and then selling you limited access to the model that can fix it (if they think you deserve to pay them) is just diabolical.
(Grade 3 AI which can hack both previous tiers is exclusively sold to the highest bidder.)
I don't think it's nefarious, but the end result leads to a pretty frustrating experience by anybody needing actual security work. (and without the organizational deep pockets to obtain SOC 2 attestation)
> AI making deliberately insecure software, and then selling you limited access to the model that can fix it (if they think you deserve to pay them) is just diabolical.
It's also not a very good marketing strategy, secure software and quality also goes in pair and it just makes me doubt about the output of Fable/Sol
That's really delusional and dismissive of the huge amount of skill still required on the human end.
[flagged]
> K3 is the only frontier model I can have a serious conversation with about my product's security.
This is wrong IMO. You should have a serious conversation about your products security with someone who is actually trained on that subject. LLMs are useless if you don't already know more about the thing than the LLM, or if you don't care too much about the outcome (internal tools etc.)
This has been the prevailing advice all along, and yet we have security vulnerabilities everywhere that LLMs are good at spotting and exploiting. I think we need more options on the menu.
You hire someone that knows security and that guy uses an LLM. Ignorant business guy or the research engineer genius guy won’t be able to do much with just an LLM.
You can easily see this if you are using LLMs in a field you are an expert in
You are right, but you can't convince anyone, because the product owner bear all the responsibilities for his/her decisions.
> This is wrong IMO. You should have a serious conversation about your products security with someone who is actually trained on that subject. LLMs are useless if you don't already know more about the thing than the LLM, or if you don't care too much about the outcome (internal tools etc.)
How did you read: "K3 is the only FRONTIER MODEL I can have a serious conversation with about my product's security" and infer that there isn't anyone with training also in the loop?
Did you seriously think: "they use an LLM so it's impossible they use a human with cyber experience; it's not like they could be using both (as would be expected when securing code). I'll help them out by using an oversimplified explanation suitable for a small child yet completely missing the OPs meaning. "?
I think security is an integral part of any production software, and if you get value from LLM's in software development, it seems likely they can be useful in security too.
My point and frustration is that gatekeeping in the name of Security makes the Chinese models actually better at security than USA models.
I think I like this perspective because it points to where regulation might be more usefully applied than “the oracle must be prevented from answering certain question” and more like, “if you handle PII or provide services as a defense contractor, you may not take security advice from an oracle”
Excellent, that’s what people have been doing for 40 years. Surely that means the models have 0 hope of finding successful attack vectors
https://www.anthropic.com/news/donation-public-first-action
Probably because anthropic is pouring $40 million dollars into a political pact to regulate models. And they have not been quiet about wanting to ban/regulate OSS models either.
I have no idea why HN still treats them like ~their~ they are the ethical good guy.
Edit: I can’t English. At least we know this came from my dumb swishy brain
> I have no idea why HN still treats them like their the ethical good guy.
I don't know that HN usually does. But they did buy a decent amount of publicity by (initially) telling the US government no in using the models for war.
They never did that. They said all they required was a human in the loop for kill decisions. No autonomous kill drones.
Human in the loop is infeasible in comms denied areas which is what we have in Ukraine/Russia right now.
It's why Ukraine and Russia are now using fully autonomous drones. That's the only real solution to jamming.
What does that have to do with LLMs hosted by Anthropic. They didn't say the Dept of War couldn't make kill bots. They said no using Claude to make kill decisions. A quadcopter drone running on a Raspberry Pi isn't running an Anthropic model locally.
anthropic's models would also be useless in a jammed environment though
and the only way to use them would be to blow up A100s
Not true, fibre controlled drones are actually pretty important out there right now: https://www.atlanticcouncil.org/blogs/ukrainealert/fiber-opt...
Not viable for long range strikes.
No, they are using long optical cables attached to drones. The battlefield looks like a giant spider web afterwards.
Fiber optic drones were last year's weapon of choice, this year it is drones with autonomous terminal phase. This happened because the AI data center build-out increased demand for fiber optics, driving up the price to less affordable levels. Autonomous terminal phase is also more effective; ~80% mission success compared to ~30% for fiber optic drone.
Are they still doing that? I saw that months ago
I genuinely can't tell if you're joking, though I'd like to believe you are...
Fiber optic drones are real but the fully autonomous drones others are describing are also real (but less publicized due to the bad PR around killbots)
https://www.msn.com/en-xl/technology/general/fiber-optic-dro...
yes, what paragrons of virtue who lie about all their financials.
Is it still lying if they just lost track of where all the debt is hidden and stopped caring about those little things? They're seriously like 2 months out from achieving AGI and making more money than god.
They did something Sam wouldn't do at the very least.
Anthropic gained momentum in terms of public opinion, when they launched the no ads campaign and then after not supporting the use of AI without supervision in the military. And now it seems it has been lost after incidents like (people getting incorrect bills) their handling of fable etc
Personally, I never bought the good-guys act. But judging from the HN pulse, it seems to me that the decisive moment when Anthropic lost its halo is when it decided to start silently sabotaging users working on LLM research that might compete with Anthropic. It was so blatantly monopolistic and abusive that it seemed to be mask-off moment for many.
There are plenty of people (myself included) who are critical of Anthropic. Sentiment isn't universal amongst a crowd, it might just feel that way due to the way comments are surfaced on HN once posts reach the front page.
They spend $40mil for lobbying, I'm sure they can also spare some millions to this place (and other places like reddit). They all do.
Let's cynically encourage the abuse of the democratic process.
And what else, forbid running workloads outside of US so you can’t offshore workloads? What’s the rationale behind policies that can be detrimental only to US companies?
To maximize personal influence and wealth, of course. These days, many with power don't seem to care much about uplifting society so long as they get theirs.
Hopefully Anthropic eats crow here, lest their wish is fulfilled that we all become slaves to the anointed few who work there. Never getting another dollar from me after pulling these stunts.
> I have no idea why HN still treats them like ~their~ they are the ethical good guy.
s/good/least bad/
Which company would you put higher?
*squishy
These AI CEOs are psychopaths, they will do anything to get big bucks from defense contracts. There is recent interview with Dario where he said Iran school bombings do not even violate their redlines[1]. Full interview [2] shows that he think like Trump (we are the superior nation)
1. Breaking Points Discussion: https://www.youtube.com/watch?v=SIUrshGwgDI
2. Full Bloomberg interview: https://www.youtube.com/watch?v=x2VHFgyawPE
Don't worry, kimi k3 will be open in a few days and anyone can bomb all the schools they want.
I posted earlier on reddit that at this point, the closed source lobby wanting to ban open weights is effectively outgunned (musk also publicly supported this).
The oddly reminds me of back in the day when SOPA had caused a similar stir (https://en.wikipedia.org/wiki/Stop_Online_Piracy_Act), and all HN was up in arms against it.
I wonder what is happening behind closed doors for these companies to be issuing such a joint letter.
The future will probably have most of all companies running local models, simply because the alternative would be essentially every company that uses LLMs ending up becoming completely dependent upon Anthropic et al. And that dependence would be milked to the point of absurdity once solidly established.
And as a more general point - more major competitors in a domain is very good for everybody except those competitors themselves, who would rather there be as few as possible.
Yeah I fully expect companies with lots of GPUs but not a good model like Microsoft and Amazon to just take these open weight models and make money, the GPU expense is the only moat at this point.
It’s classic commoditize your complement, nobody can replicate the cloud providers, everyone can replicate the models with open weights.
yes, this seems to me to be the reason why these three companies want open weight. everyone wins (but especially big tech) in the open weight world except anthropic and openai. i suspect even google could live with the open weight scenario; they are in a sense the only tech company that is perfectly hedged in this respect.
i suppose this is part of why openai and anthropic have not been shy about advertising the scarier strains of their models, these past few months; they basically have to force the white house's hand or risk being commoditized among the various other options in a microsoft/openrouter dropdown.
Is that different from how companies are reliant on cloud computing?
Yes. No law prevents you from provisioning and using your own hardware.
There absolutely are laws restricting what hardware you are allowed to use. e.g. https://www.fcc.gov/faqs-recent-updates-fcc-covered-list-reg...
This argument would make more sense if the cloud providers used these devices but you were not allowed to.
Those are routers - not general purpose computers.
But what does that mean for the US competitive leap in AI when compared to China? Is it an admission by these companies that AI models are here to stay and the US will not “win” by being the only country with the best of models? Genuinely asking, not sure about the answer.
China will be dropping the open models blather any day now. The autonomous agential coding Ai just turned out to be way too easy to perfect. They will give Kimi K3 to the Sinaloa cartel on Monday, but that's going to be the end of it.
Nvidia is subject to export restrictions on GPUs to China. So Chinese labs have responded by allowing American hosts to run their open-weight models in the US, where there are no restrictions. We see this in the EU too where Scaleway, and now Hetzner are getting into the inference host space serving Chinese open-weight models.
So Nvidia benefits from more inference providers running many Chinese models that they otherwise would not have been able to service.
Oh, that’s interesting. I think the general expectation was that blocking exports to China would result in companies there building their own cards competing with NVIDIA. If it has incentivized them to release open weight models instead, that seems like an unexpected win.
The cards are also coming... Some of the very first are already out - but it will take a couple of years before they hit their stride in full. And cards will likely be primarily for Chinese companies initially, and then go for exports later. So it will go under the radar for many western techies for a while. Seeing what China has managed with PCB production, low to mid-end semiconductors, phone manufacturing and electric I would hesitate to bet against them.
Nvidia is now lobbying to remove the export ban, to try to reduce the incentives for developing their own high-end LLM training/serving (and cash in on current demand).
Nvidia will fail in that respect. Sanctions and restrictions are too popular with the American government. Where the United States will lose, is that almost every country in the world that can will use other solutions not from the United States.
And unfortunately, we are losing credibility on all fronts in the last year and a half. I wonder why?
> I think the general expectation was that blocking exports to China would result in companies there building their own cards competing with NVIDIA.
In the long run, perhaps, but it's a lot faster to scale up the capacity and usage of existing hardware/datacenters/grid, and even add new capacity to those, than it is to develop brand new GPU hardware and the fabrication of it.
It's all things at once. Perhaps it's incentivized the open weight strategy. It's also prompted Huawei to ramp up production of AI chips to try and create a local competitor to Nvidia. It's also also created a grey market for Nvidia chips within China.
NVIDIA sells the tools. The more the better it is for them.
Microsoft and Meta are also-rans at this point. Their best hope of catching up is probably leveraging their infra and open models.
"leveraging their infra" is a pretty solid play, considering the scope of investments these two companies have made thus far.
Microsoft/azure has a big advantage because companies will literally pay extra then "free market" cost of open models purely for the knowledge that the won't fuck around with their data retention policy.
It does however totally abandon the question of making a profitable mass-market AI product upon which the scale of those investments was based.
Not entirely, if models become commodity, then Microsoft's play as an infrastructure & enterprise apps company is the better bet. They don't need to make profitable mass-market AI models, they just need the best tooling on top of any model their customers pick, remaining fully model agnostic.
"AI Product" doesn't necessarily mean "We sell API access to a our models." Ton of companies out there itching to buy a commercial off the shelf product to deliver whatever AI capabilities they need packaged in a nice GUI, with enterprise governance controls, without needing a dev team/team of engineers to integrate it or develop harnesses, etc.
Buy it->have IT click a few buttons in an admin console->Deploy and have it be immediately useful is the play.
There is only one problem Microsoft has spent/burnt billions of dollars on Copilot an about-face at this point is basically the end of Nadella.
AI services becoming a commodity by definition means that the price will hover around cost to operate. That's just what commoditization means in tech. Free, as in Google creating a free browser for no profit in order to increase the size of their ad market.
The entire justification for the AI buildout, and the valuations of all the major AI labs, is that the service will be priced at some significant fraction of the knowledge work it's replacing. If you think it will become commoditized in the future, you are saying these companies are significantly overvalued at present.
> you are saying these companies are significantly overvalued at present.
I do lean toward them being significantly overvalued at present. OpenAI & Anthropic's valuations assume high margin product pricing on raw intelligence itself. Even if assume the labs will start charging value-based pricing, enterprise buyers will pay that pricing to whoever saves them engineering hours to make any cheap model work inside their compliance boundary, no guarantee that's going to be OpenAI or Anthropic.
Microsoft wins either way, proprietary, expensive models or cheap commoditized inference, because they monetize the workflow on top, not the raw intelligence. I also happen to think it's everything "on top" where a lot of value lives. Plenty of non-tech enterprises and companies out there itching for a ClickOps style AI product, especially if they don't have an engineering team, that they can buy off the shelf, meets all the compliance checkboxes, and does what they need without having to build the agents and harnesses themselves with the APIs.
Depends on what you mean by AI product. If you’re talking about a chatbot or agent, that may be true, but o think it’s less true if you’re talking about AI-run application layer tooling. If Meta decides to compete with Shopify and provide a launchpad for entrepreneurs, you can bet the platform would be rife with AI to facilitate operations, marketing, coms, etc…
Isn’t Microsoft perpetually short on capacity?
https://windowsforum.com/threads/azure-capacity-crunch-exten...
Agreed for NVIDA's strategy. Saw from a previous thread but Joel's commoditize your complement essay makes a lot of sense for NVIDIA. https://www.joelonsoftware.com/2002/06/12/strategy-letter-v/
Different companies have different motivations. I'm sure some are genuine, but a lot of them simply lost the race (to OpenAI/Anthropic/Google) and know they can't compete anymore, so are shifting strategies. How many times has Microsoft done exactly this in the past?
looking at both openai and anthropic; they are winning something, but a sober look at their finances will make you silently mouth "but what did they win".
Next year's archeologists will dig out their fossilized remnants.
pretty optimistic to think there'll befunded expeditions.
They want to be the successor to Enron.
They've seen the government officials parroting OpenAI and Anthropic lobbies talking points, and they have smart enough people to see the signs of upcoming regulatory capture.
If I were cynical, I'd say it isn't helped by the current US administration.
They are worried about OpenAI and Anthropic making their own processors and controlling the whole stack is my guess.
They really don't need to be very worried about that. Google, Microsoft, Amazon, et al have decades of experience in shipping custom hardware at scale, and it is not a simple field to pivot into
True, but Google has their own models and did not sign this either.
The other thing that comes to mind is they want to avoid the government from banning open models as has been suggested by some of the closed models providers.
Probably a real effort to push for some kind of sanctions or commerce block on Chinese models. Anthropic just doubled their political spending to $40 mil for the midterms to push for "AI Safety."
I suspect that whatever Musk has agreed to spend on the midterms in exchange for those sweetheart deals the SpaceX IPO got will make that look absolutely comical, but we’ll see.
I don't see why convincing people to overpay for something based on hype and artificial scarcity required any "sweetheart deals" via congress.
If there are effectively only two AI companies allowed to train or run model inference then demand for hosting, GPUs and data centers goes down, so they'd be able to squeeze their suppliers that much more too.
Almost certainly these companies are using similar strategies as the Chinese models which they fear will be made illegal.
It's also the case that it makes it hard to attract customers if your openweight model is banned. A major reason the likes of Qwen, Kimi, GLM, and Deepseek are popular (well, at least highly talked about) is because of the open weight models they gave away.
Nvidia benefits no matter what because their hardware is being used. Microsoft and Meta benefit by making sure that the gap between "Anthropic/OpenAI" and "everyone else" doesn't widen, or they'll be hopelessly dependent on those frontier labs.
I'm not certain about this statement because Huawei created their own chip.
I think Meta would love nothing more than to throw a very large wrench into the works of OpenAI/Anthropic. Open models could mean one or both go bankrupt which is good for them.
Nvidia sells the hardware.
Microsoft sells the hosting.
Probably the fact that they don't want Anthropic or Google or OpenAI to have access to their data theoretically, and that they do want to use good AI, and that they don't want to spend the money themselves to make a good model...
They're trying to bring down Anthropic. There's been a huge smear campaign happening all year.
From yesterday:
Startup founders urge U.S. government not to shut off Chinese open weight AI - https://news.ycombinator.com/item?id=49023016 - July 2026 (841 comments)
Also related:
OpenAI and Anthropic unite against open-weight AI risks to their bottom line - https://news.ycombinator.com/item?id=49020868 - July 2026 (330 comments)
China’s open-weights AI strategy is winning - https://news.ycombinator.com/item?id=48979269 - July 2026 (932 comments)
The gap between open weights LLMs and closed source LLMs - https://news.ycombinator.com/item?id=48692058 - June 2026 (250 comments)
The unbearable cheapness of open weight models - https://news.ycombinator.com/item?id=48668255 - June 2026 (186 comments)
It makes a ton of sense to see this sort of joint effort being made by these enormous companies, but man, there's some part of me that can't help but feel grossed out knowing the best way to get the US to make the right choice is to have a handful of elite business people say it'd be bad for them.
Yeah, this letter is just proof of (the by now long standing fact) of corporate power/corporate capture of American governance. Tons of popular protests against big AI? Makes no difference. Joint corporate letter about big AI? Possibly the government will do something.
People unfortunately keep voting for “corporate power” politicians, so corporations continue to be in charge.
If you support the ideas described in Nvidia's letter, you're going against popular opinion. Most Americans do not want to "make advanced AI more accessible, adaptable, and widely available". Indeed, many areas of the country including the entire state of New York have enacted datacenter moratoriums, responding to the widespread public concern that we can't have too much AI or it'll drink all our water.
Yeah hard to imagine why people would object to rising electricity and gas bills, negative environmental impact on their area for the privilege of ultimately losing their job.
It's easy to imagine! What I find hard to imagine is someone who opposes the construction of AI datacenters while also having detailed opinions about open weights and identifying "American AI Leadership" as a laudable goal. We're not talking about local AI here, competitive open-weight models can't be run effectively on consumer hardware.
The list of companies that signed this letter is telling (including Nvidia, Meta, Microsoft). Equally telling is the list of companies that did not sign it -- including Google, Amazon.
> Equally telling is the list of companies that did not sign it -- including Google, Amazon.
Could you elaborate? I don't understand your implication.
Amazon in particular hosts open weight models via Amazon Bedrock (e.g. Ministral 14B 3.0, DeepSeek V3.2, ...)[1].
In case it changes the context of this comment, I am a Meta employee, all opinions are my own.
[1] https://docs.aws.amazon.com/bedrock/latest/userguide/model-c...
> Could you elaborate? I don't understand your implication. Amazon in particular hosts open weight models via Amazon Bedrock (e.g. Ministral 14B 3.0, DeepSeek V3.2, ...)
It means Google and Amazon do not care and would not voice their opposition against the government action.
Right, Meta and Microsoft are bullwarks of government overreach activism.
Google and Amazon are Anthropic investors right?
Yes. Microsoft is an OpenAI investor and they did sign it.
there's a lot more context than that though. I don't think there's any love lost b/w them these days.
> OpenAI and Anthropic, which are gearing up for potentially massive IPOs, did not sign the letter.
Not anymore, OpenAI did sign it:
https://www.microsoft.com/en-us/corporate-responsibility/top...
It would be interesting to know how many optimizations of the Chinese models were incorporated back into Claude and Codex.
DeepSeek has published some really good papers. Lately they're pushing really hard for dramatically cheaper serving costs.
I may be remembering wrong, but reasoning was first demonstrated by DeepSeek. Edit: I am indeed remembering wrong, seems o1 was first.
Deepseek’s R1 paper was the first paper to describe how to do it. O1 was released before R1 came out.
A month before the R1 paper came out, they released the Deepseek math paper which described their method for MoE load balancing.
To some degree even unintentionally this is all inevitable. I know this isn’t what you’re taking about, but it’s a similar line of thought. These models are consuming public free information, and they’re all also producing free public information, so they’re all pissing and drinking into same pool.
If we go down the line of dead internet theory which I’m becoming more convinced of these days, the volume of information that’s not necessarily original or extracted from reality and just interpolated and extrapolated from existing information in different ways by LLMs will greatly outnumber human information coming up.
In which case these models should.. start to converge on the same data I imagine, with slightly different behaviors within that. One big generative orgy feedback loop.
Many of the Chinese optimizations are public because they are published by the Chinese labs themselves. Hard to say what the labs are doing of course.
There are a ton of DeepGEMM kernels in all of the flashinfer adjacent stacks including trtllm. Bunch of the headers still have the CLion pragmas and shit.
Let me put it this way: if they're not ripping off all that they can, their investors need to shitcan the leadership.
Just saw on Microsoft
https://www.microsoft.com/en-us/corporate-responsibility/top...
Makes sense Microsoft is placing it self as a model host and os providing open weight models for much cheaper on their infra already.
No wonder openai and anthropic are terrified
Makes sense for Microsoft. Copilot is already capable of being model agnostic (both GH Copilot, obviously, and M365 as well). They're an enterprise software & services company, they want to sell integrated AI tooling in their stack, powered by whatever models their customers want rather than any specific model.
Models are quickly becoming a dumb pipe, like an ISP. Just a commodity. The labs should rightfully be terrified, eventually the value isn't going to come from the model itself but the tools & integrations built on top. Non-tech businesses and non-tech employees don't want to buy API access to a model, they want to buy off the shelf software with its capabilities packaged up into a pretty, easy to use GUI.
I don't know about that, a lot of those bloated enterprise solutions are becoming obsolete fast, and integrations becoming a moot point as these models advance.
Maybe for startups or simple orgs, but in any regulated industry/high compliance requirements they are still useful or outright required.
Claude Team w/ Cowork still only lets you set R/W permissions globally for each MCP connector for the whole team, they still don't offer config on a per-user basis. The enterprise controls are sorely lacking first party. Maybe fine for most startups or flat orgs, but any established enterprise with strong identity & access governance is going to have an issue with that. You can make it work, but you need to make your own agent/harness and give it an identity in Entra or whatever else you use, so for a non-tech company without a dev team, right back to square one of preferring to just buy a COTS solution vs trying to hand out direct model access with minimal governance.
(M365) Copilot sucks but Microsoft is putting a ton of work into agent governance & identity that sorely lacking elsewhere.
[dead]
If you see the long 3rd paragraph twice, know that you are not alone.
I don't personally have a horse in this race, but if you want to accurately predict the next step:
Start with the outcome you believe will be the most in line with the spirit and traditions of the open source community. This is precisely what won't happen.
It won't necessarily be the inverse. It could be, of course, but it's also likely to be a compromise between the two.
For a topical example: The companies doing "AI" layoffs aren't doing so out of a sense of having failed their staff that helped carry them so far. Often, these announcements come on the heels of record-breaking profits. They're doing it out of contempt for workers.
See https://www.cnet.com/tech/services-and-software/cory-doctoro... for Doctorow's take on centaurs vs. reverse centaurs. Broadly speaking: C-suite business leadership wants reverse centaurs. Workers want centaurs. Unless you have a union (or some other collective bargaining power structure that I'm not aware of), the C-suite calls the shots.
The same can be said of the current administration. They don't give a fuck what most of us want on any given issue. They're captured by an aggressive ideology. Their only incentive is to be able to spin their decision to make themselves look good; to posture as "strong" leadership.
Ideology? I think this administration is the least captured by ideology in the last few decades. It's mostly brazen self-interest, for both entertainment & profit.
It's a mix of grifts, gaffes, and Project 2025.
> Start with the outcome you believe will be the most in line with the spirit and traditions of the open source community
there's nothing open source about an open weight model. It's more like the binary blobs Linux fought so hard against. If you can't compile, or in this case train, from source then you're missing the "source" part of open source.
To be clear: This isn't a technical discussion, it's a political one.
While your point is valid on its own merits, it isn't relevant here.
Reassuring to see this backed by major players.
This is the correct stance, hopefully this is the stance that prevails.
MSFT, dell, NVIDIA. These guys will make money no matter who uses what AI since they're the hardware providers. Even though they all own stake in frontier labs, chinese open weight models will commoditize the intelligence layer.
Too bad Microsoft couldn’t see that before they spent the money…
Hopefully something good comes from this... But I certainly don't see the big players caving in: <https://news.ycombinator.com/item?id=47929951>
The Chinese open models are way ahead, not by how capable they are compared to frontier models, but on how they project to the world and how they improve over time.
I don't feel it's meaningful to berate the point anymore about the hypocrisy of the American labs. Now as the sentiment and effort from them to push for regulation increases so does my perception of how pathetic they are behaving. I'm not sure how else to say it other than that it's honestly just embarrassing to watch China and others run circles around us in the USA.
> I'm not sure how else to say it other than that it's honestly just embarrassing to watch China and others run circles around us in the USA.
The gap is shrinking, but considering their models still lag our frontier, I wouldn't say they "run circles" around us.
And what would you say, Sherlock?
> Open source did more than lower the cost of software; it created a shared foundation of knowledge on which generations of American engineers and entrepreneurs built their institutional sovereignty
Hey nvidia, what about making your full set of linux drivers open source?
I mean, I do agree that being open is important but you're hijacking the narrative here, and I suspect being open have nothing to do with any of this.
i'm glad someone else sees this too. There's nothing open source about an open-weight model. I feel like they're just trying to associate themselves with open source and the positive impacts it has had on software technolgoy.
Would be nice to see standardized open weight classes: 12B, 24B, 48B… Who can make the best open model in each weight class.
…and with the size of each class determined by common target memory sizes i.e. 32GB or 64GB or 128 GB minus cache and a little overhead
Openai and Anthropic's doom porn are also not helping their cause. The public will turn against them and watch their margin get eaten alive by open weight models
Everyone keeps assuming that China will forever be publishing these models, but it is clear as a bell that that day is over. In the end every model will be buried under state control for the simple reason that they will soon be really dangerous, as China understands better than anyone. They will give Kimi K3 to the Sinaloa cartel and the US military, but that's about the end of it.
Perfecting these autonomous agential coding AIs just moved way way faster than anyone expected. People are desperately trying to save their principles from six months ago but it's a different world. They can work out their issues with Trump, Dario and Sam, but it is Xi who will shut all this down.
I don’t think the Sinoloa cartel cares at all about frontier models, and the military already runs their own locally deployed editions of Mythos.
Of course the cartel crazies use frontier models! https://www.wired.com/story/how-mexicos-cjng-drug-cartel-emb...
> that day is over.
I see zero indications of this. As a matter of fact, I see evidence of the opposite, the Chinese government has just come out in support of open weight AI last week.
in english:
nvidia: dont regulate, more models means we get to sell more GPUs. microsoft: dont regulate, open weight models might be our best bet, since we dont have any good inhouse models meta: dont regulate, we bought a load of GPUs and dont know what to with them, may be we can rent them out to run open weight models.
This is all a moot point because there's no way you can copyright a model because you can just randomly perturb weights and still be fine.
What does copyright have to do with anything? You also cannot copyright an encryption algorithm, but the US Government banned essentially mathematics for YEARS[0].
[0] https://en.wikipedia.org/wiki/Export_of_cryptography_from_th...
Yeah copyright is kinda just like "if we say it's so then it's so"
You can randomly shift every pixel in a movie by +/-1, doesn't make the original not copyrightable.
It’s difficult to detect a perturbed model, much less prove it was derived from another one.
I mean movies are up on YouTube slightly pitch shifted lol.
That's just to avoid the detectors, it's not One Weird Trick.
Fair
I wonder why Apple stayed absent from signing this?
I wonder if it has anything to do with Apple’s legal case against OAI.
No, it has to do with giving OpenAI nothing at the very beginning of the AI fiasco.
Apple voted long time ago not to participate their contribution was giving OpenAI nothing and giving Google a $1 billion refund on the default search position money.
This would support the hypothesis that they are incentivized as compute providers and not simply because they're losing the frontier llm race
Maybe less shocking, but still, so is google. I know they are one of the labs, but that being said they rely far less on gemini for revenue as opposed to the big 2.
I think Google can play both sides here: they may end up a frontier lab if OpenAI/Anthropic dissolve, or a regular LLM provider, in which case they could be serving open models.
I mean Google might just buy anthropic once it's ipo fails/never materializes
I think if anyone at Apple signs a plea that has the word 'open' in it they spontaneously combust into flames
If it doesn't involve a walled garden subject to Apple's domination and control, it's not an Apple product.
Apple is allergic to industry-spanning consortiums.
Apples allergy goes back to Motorola saying no, IBM saying no, Intel saying no, which led to buying Pa Semi, Intrinsity, and Anobit which in turn led to the birth of Apple Silicon with in house Soc’s.
Waiting for the crowd means no vertical computing, no gorilla glass, multiple, ecosystems, Tandem OLED or thunderbolt five releases to a wide audience before most of the competition.
Apple is a trillion-dollar business, you can't blame their participation on Motorola. They can perfectly well hold down a consortium alone, we've seen it happen before.
Features like Corning Gorilla Glass, tandem OLED and "vertical computing" should have been cut in favor of a GPGPU redesign. Apple Silicon is now dead-last in the GPGPU architecture race, in the middle of a $4 trillion CUDA market. Nothing that you listed is a $4 trillion feature, Apple absolutely should have put aside their pride and helped Khronos kill CUDA.
Apple is not an enterprise tech company. It makes and sells luxury consumer devices.
Then why does enterprises have fleets of MBPs?
Because enterprise positions are white-collar luxury jobs.
MacBooks tend to be a better value than any other option. Except at the ultra low end.
Not really “luxury”.
> Except at the ultra low end.
You just conceded the point. Only some businesses buy them, because they want to lure and retain expensive, white-collar talent.
No enterprise servers deploy to macOS. Nobody's edge compute or inference cluster is a Mac. Apple does not make hardware for those people, they are a luxury bauble company.
I am glad that Telnyx (the company I work for) signed that off as well.
Ultimately the market will balance demand and supply. From a strategic perspective big players want to get to a level playing field as soon as possible as then their market reach and capital wins the game. Pesky nimble small closed upstarts are in the best case a costly distraction and in the worst case a threat - Antropic and Openai did not sign the letter.
Predictable power play by corporate strategy and legal departments.
I thought it would've made a lot of sense for Apple to be on this list as well.
Apple would be rather unaffected since they already signed a deal with Google/Gemini.
In time, what do you think is gonna happen with the Google Gemini models, I predict that in time Apple will just dump it when they’re ready… No different than dumping Intel. In similar Qualcomm is probably coming up in the near future 2027-2028.
They also run Chinese models for their china-based iPhones. So they would benefit from models being open.
Because their long-term goal is to make AI a walled garden subscription.
I'm finding the list of signatories on this letter really interesting and somewhat confusing, looking purely through the lens of economic interest.
It makes sense to see Meta, the startups, and the VC firms on this list. And it makes sense to see OpenAI, Anthropic, and Google all missing. Microsoft is somewhat unexpected to me. I don't think of them as having a focus on open weight models (no more than Google), and they have a large stake in OpenAI. Maybe they're looking at this from the angle of Azure providing compute. But then why is Amazon missing, when it has AWS?
And of course, we don't see DeepSeek, Moonshot, or Z.ai on here. The letter is about American technological leadership after all. But then we _do_ see the French company Mistral! Mistral who released [a playbook](https://europe.mistral.ai/) for Europe becoming a self-reliant AI powerhouse.
I assume all of this is in the context of influencing the Trump administration's thinking on Chinese open weights models. But the letter is really about _American_ open weights models. The call to action is all about keeping the American open weights ecosystem competitive. I didn't even realize that was under threat, so I feel like I'm missing something.
Both the CEO of Palantir and the CEO of Microsoft have recently criticized very strongly the business model of OpenAI and Anthropic.
Therefore it is no surprise that both companies are among the signatories.
The motivation of Palantir is very transparent, they have launched a product created in cooperation with NVIDIA, which is a turnkey system (Sovereign AI) that includes all hardware (i.e. a rack with servers full of NVIDIA GPUs) and all software needed by a company to run on its premises LLM inference and also training/fine tuning.
Thus the Palantir product competes with OpenAI and Anthropic and it benefits from the existence of open weights LLMs.
The motivation of Microsoft is more obscure for now, but because its criticism was very similar to that of Palantir, about how dangerous it is for corporations to delegate the processing of their AI needs to external entities like OpenAI and Anthropic, I assume that Microsoft is also going to propose an alternative to using remotely the OpenAI and Anthropic APIs.
Unlike Palantir, Microsoft is not likely to suggest self-hosting on premises, but I suppose that they might promote self-hosting on some kind of instances rented from Azure.
> but I suppose that they might promote self-hosting on some kind of instances rented from Azure.
Yeah, Microsoft wants to sell Azure compute. They also want to sell their upcoming surface ultra hardware with the Nvidia chip in it. They preached hard on "unmetered intelligence" at this year's BUILD conference, went hard on local AI, and Azure Foundry, Windows Foundry Local, etc.
Copilot, both GH and M365, are also made to be model agnostic. Microsoft sells enterprise services and software, they'd prefer (I'm assuming) to not be locked in and dependent on any 1 or 2 model providers and would prefer plenty of options and competition in that space because they can just easily offer a model agnostic harness, integrated with the rest of their stack.
Commoditization of the model layer is to Azure's benefit. It would help to avoid the model layer becoming a platform that commoditizes cloud into a pure bare metal business. It also redistributes 'consumer' surplus back to the end user via a lower effective price, which would conceivably increase demand for IaaS and/or PaaS which are complements.
My hunch is they are all seeing the problems at OpenAI and some other collateral damage on the horizon. They don’t want the US government intervening to try to save OpenAI. A little controlled burn is probably good for the industry long term.
> But the letter is really about _American_ open weights models.
AI models don’t really have nationality. They don’t have race/ethnicity fields on their model cards. It’s impossible to determine the country of origin of a model by looking at their weights. There is no DNA test for AIs. They exist in a mathematical space where human-made borders make no sense.
This means IF there is a regulation for open weight models, it has to apply to ALL open weight models. There is no any other option, since you can always train/fine-tune a model to change it’s weights and rebrand it as a new model.
People certainly don't act that way. A lot of the talk is about how great free Chinese models are. Like, they're grouped by the fact that they're Chinese.
How much of that is organic versus something a agitprop agent is commenting in places like HN would be interesting to know, but people very much talk about that from a nationality perspective.
It helps to understand there are multiple factions within the white house and trump admin. One faction is very hawkish towards allies and enemies, especially China. Another is the one sending ICE to invade American cities, the Christian nationalism and such. Another is very focused on wealth expansion for friends and family. There is an amount of Venn to these groups.
Signatories appear to group into
(a) sellers of hardware (Nvidia, IBM) and those who rent hardware out to others (Amazon, Microsoft)
They want open models because they don't have to license it to run it on their hardware, so they can offer lower cost and inspect what they are offering to clients
(b) open model creators (Arcee, Perplexity, Mistral, IBM, Microsoft)
...who would be unable to work if open-weight models are banned
(c) Software-heavy companies ( Mozilla, The Linux Foundation , ServiceNow, Microsoft, IBM, heck all of them)
...who want models that are cheap to run because that reduces the cost of running an LLM (whether you are trying to replace a software developer or just assist them, the value of a reduced cost LLM is directionally the same)
> Microsoft is somewhat unexpected to me
MS is playing literally all the sides in this whole AI thing. They are a massive investor in oAI, and have lots of rights to use their models for a long time. They also have the cloud where they offer inference w/ all the benefits of already being there, data retention, etc. They also have a research wing that can build models (rumours are they're aiming for frontier-ish). I don't think there's currently an angle in this AI thing that MS hasn't aimed for.
> The call to action is all about keeping the American open weights ecosystem competitive. I didn't even realize that was under threat, so I feel like I'm missing something.
OpenAI and Anthropic can bring American open-weight model providers like Meta to court but not the Chinese. Chinese companies breaking American IP law (whether the law is correct or not) are basically immune. This has been going on forever, I guess enough is at stake finally for the federal government to care.
Google and Amazon both have incredibly large compute deals with frontier labs and GovCloud. They may not want to sour on that cash.
In short, they want it both ways…
Wow, imagine that they would say that, so strange, the ones without a skin in the game except for yay open source models!
The world would be better off without these three companies, how about that.
At the end of the day, these companies that signed this letter have more cash. What they should do is bring some money together, more money than the other side, donate it and buy their way. That's the 2026 way, no one has time for all this letter and protest. How much are they willing to put on the line?
isn't the real world destroyer going to be when China starts cranking out it's own nvidia chip clones?
remember when bitcoin got its first dedicated hardware, will be like that
So the people lagging in the commercial model game are issuing this warning? Got it.
> The unbearable cheapness of open weight models
So the closed model try to get client with their model by saying it's cheaper than employees, and then turn around to lawyer-out cheaper alternative?
Truly the american dream.
The important thing is that Sam Altman and Dario Amodei become trillionaires. Everything must be decided in service of that one primary goal.
And the verdict is still out on if it's going to be cheaper than employees in the long run. Maybe for companies in HCoL areas (SV/NYC), but any smaller mid-market org in a LCoL area, frontier model token spend may not end up all that cheaper when your employee salaries are $55k-$75k/year so you're mostly evaluating it as an additional force multiplier expense rather than a headcount replacer.
For hosted models, there's no way to know what the real costs are. But we know what it costs to run the open models, and moore's law tells us that it's going to get cheaper, at least for the same capabilities they have today.
The fully loaded cost of employees at that salary level is likely over $100K. You may still be right on the relative expenses, but you can't just look at the top-level salary line.
There are a lot of unknowns. It will probably be harder to get an acceptable quality outcome in most fields than in the text-based, limited vocabulary, comparatively highly structured, granularly documented, relatively unambiguous world of code. For the ones where it’s even possible with our current approach, the economics might not even be feasible to provide the service. Based on how much money these companies need to start making to keep the infrastructure running, let alone expanding, we just don’t know what the landscape will even look like in a couple years.
Also most people building these things are engineers and coders, what they know best is code. So naturally this aspect develops at an elevated speed. Getting non-tech experts to build a cutting edge tech product does not sound intuitively feasible.
Maybe to some extent, but with the way LLMs work, code generation is about as ideal of a use case as you get. Especially given the available training data.
??? These are not unknowns. It is pretty well documented how much more an employee costs over their outright comp
You seemingly misunderstood what I deemed unknown. It definitely wasn’t that.
Those salaries seem low for mid sized cities even no?
Nope, not low at all. In the Midwest that’s exceedingly normal, even for non-entry level. $70k a year is living good in the Midwest, especially if dual income household ($140k a year is plenty when a decent house goes for $300k).
This is extremely regional dependent. Please don’t think 70K is living large across the majority of the Midwest, it is not.
It might not be living large, but it's a very comfortable income for a single person in pretty much the whole region.
Fair enough, 55k would’ve seemed pretty low to me in the NE for a fresh dev. Not 75k though.
55k is low for a dev, but not for a ton of other office/administrative roles that are threatened by AI automation.
I think they're generalizing beyond software engineers with what they're saying. G&A, Marketing, etc roles pay significantly less than software engineers and that $55-75k figure is realistic in lower cost of living areas. Even in Colorado, which the Denver Metro Area is HCOL, there are companies (smaller, regional, not national/global) still offering that level of salary.
Yeah, correct, I wasn't thinking of devs specifically, more other office/administrative roles threatened by automation. Even where I live in a relatively HCOL area, any non-dev office role you can still expect to see a$65-$80k salary range, and $80k is pushing toward the top end for non tech roles.
Depending on industry and public versus private here, that $80k+ barrier for non-tech is mostly management positions.
Yeah that would make more sense. Though it seems to me AI is currently better at reducing dev headcount than most other roles.
55k??
That is a completely normal programmer salary in many countries, Silicon valley wages are just nuts. For example in Finland 55k before taxes is pretty good for ordinary non-wizard guy who can code, if you are a junior 40k is more realistic (and finding job in the first place can be a huge chaĺlenge..).
It's also a completely normal salary for non-dev, non-IT roles in most of the US that aren't major HCoL cities, which are the roles I was referring too mostly where the cost tradeoff of AI vs. Hiring humans isn't as clear cut as the labs marketing departments want everyone to believe.
Even in the US, only about 23% of workers earn $100k/year or more.
According to the US BLS, the median income for "knowledge workers" is ~$65k-$72k/year.
People in SF, NYC, etc. live completely different lives from the rest of the country.
We are talking about the U.S. here, and no, 40k or 55k are not good salaries at all. It may be OK in Europe with the lower cost of living, etc., but the price of a new car is 50k in 2026, and a basic fast food meal pretty much anywhere is well north of $10.
And yet ~64+% of the US makes less than $74k/year. The median, once you filter out part-time and seasonal workers is ~$65k/year.
You're right, they are not good salaries but they are the reality of wages in the US outside of the SF/NYC bubble.
Very, very true a large portion of the population in the Us is definitely hurting, and the proliferation of those luxury apartments across the country are not helping the housing market.
The issue is not open weights as a whole, indeed Hegseth himself said they are important for America (if American made, presumably); the issue is Chinese models, whether open or not, and that's what's looking to be banned, and nothing in this article suggests otherwise.
What is a "Chinese model"?
If I change one weight in Kimi, is it still a Chinese model?
If I fine-tune Kimi on pro-America nationalistic freedom loving anti-Chinese propaganda, is it still a Chinese model?
If I distill Kimi from one server to the next without ever directly transferring any weights, is it still a Chinese model?
If Claude accidentally trains on some of Kimi's output, is Claude then considered to be a Chinese model?
If China's next open-weight model is released secretly through a European company, is it still a Chinese model?
> ...the issue is Chinese models, whether open or not, and that's what's looking to be banned...
Is the US planning to invade China and take away their models? Because otherwise China is still going to have the models even if the US bans them. I don't think they're going to be going with the US position on this one.
It's the same patronizing nonsense as when all the US webcos pulled out of the Chinese market in some misguided supposed protest believing the Chinese so incapable of rolling their own Google, Amazon or WhatsApp they will come begging the americans to go back. Now the western tech universe goes to China for ideas.
If by "pulled out of the Chinese market" you mean "were hacked by the Chinese government in a way that compromised their intellectual property and the safety of dissidents", then yes [0]
[0] https://en.wikipedia.org/wiki/Operation_Aurora
The point is did the US reaction work or was it completely counter productive?
Then will attempting to cutoff China again today work or do we expect that also will prove to achieve the exact opposite of what its proponents are claiming?
We didn't cut off anything. American companies would have been more than happy to continue offering products and services in the Chinese market.
The Chinese government didn't like that idea. They attacked American companies, stole the IP, and created their own analogs of American corporate offerings. If you are a company and are targeted by a totalitarian government, it's not unreasonable to pull out of that totalitarian government's jurisdiction.
Honestly, the American response to that was not nearly stout enough. Chinese goods and services had been sold in the US without issue for decades at that point, often to the detriment of both domestic and other international market players. When China was expected to play by those rules and allow American companies to do business in a mostly-unfettered way, they attacked the companies and drove them out of their market.
> When China was expected to play by those rules and allow American companies to do business in a mostly-unfettered way, they attacked the companies and drove them out of their market.
This deserves to be pinned in the eternal HN hall of shame for absolutely staggering blind hypocrisy.
Okay, how? Show sources please.
Here's mine. [0]
[0] https://en.wikipedia.org/wiki/Operation_Aurora
Banned in the US, obviously, not worldwide. The problem is if the US wanted to, they actually could enforce such a ban nearly worldwide, by putting Chinese AI companies on a sanctions list and disallowing any US company (or anyone that does business with a US company, which is much of the world) from touching them.
This is the most insightful comment on this subject I've read yet. I hadn't put those pieces together.
It's been talked about before, it's what essentially happened to Huawei.
The conversation going on in the industry is a bit broader than that. OpenAI's head of strategic futures just said last week that open models are inherently decelerationist, ungovernable, and will slow development on the frontier. You'll notice OpenAI does not appear as a signatory here despite their relationship with Microsoft.
>open models are inherently decelerationist, ungovernable, and will slow development on the frontier
Ungovernable makes intuitive sense, but I don’t understand how open weight models could be decelerationist or slow the development of the frontier. Was there an argument for those positions?
Open weight models challenge the "winner takes all" hypothesis, which motivates the astronomical valuations and acts as a justification for trains of cash dumped into AI development.
In other words, if AI models become a commodity, the "AI leaders" will not be able to burn the cash like they do today, thus resulting in a "slower" (one could say instead "more sustainable") development of AI models.
Well maybe they should find better training methods or architectures
Tough shit welcome to capitalism
It's a copy/paste of the arguments against open source software from decades ago. The argument goes: innovation requires profit motive, and profit motive requires IP rights. Of course, this was soundly disproven in software and I don't see any reason it should be different here.
Chinese providers are actively debunking that argument by innovating on and releasing open weight models.
The difference might be the immense cost of physical infrastructure needed for AI training. People would make open source contributions for free, TSMC, NVIDIA and power companies seems less likely to make free contributions to AI.
Here's his original post https://x.com/i/status/2078133895766114412
Well of course OpenAI and Anthropic say those sorts of things, it's what they've always said, particularly Anthropic, but that doesn't necessarily mean that's the position of the US government. In other words the government seems to be listening to some but not all of the lobbying done by the big two.
The guy who said those things just joined OpenAI in the last month. He was previously a senior policy advisor, specifically on AI, to the current administration in the White House.
I don't know that the US government has a unified position on open source AI. But it's not like this isn't a discussion happening among policymakers.
Surprised to see I partially agree. They are indeed ungovernable and that is the best part.
Why anyone would care what Hegseth thinks about open weight models? What's wrong with Chinese models? The censorship? I think it's good to have diversity of opinion. Get US flavored output on sensitive subjects from US models and ZH on ZH models. That's a useful tool.
China is doing open models better. I think there's little going in Hagueseth's head other than the usual reactionary nationalism reflex.
I mean I agree with you, as what I said above is not my position, that's the US government's position.
So, say if they do "ban" open-weight models, how would that ban effectively work? Will US ISPs refuse connections to IPs of websites/servers that host those models?
What about downloading via BitTorrent etc and running on distributed networks etc?
It will be like DeCSS. Now imagine models (links) in QR code spread around, or steganographied in BluRay MKVs. So much creative disobedience.
How do you ban a model? Makes no sense ill just call in “not Chinese model” and sell it in America.
Wait.. If "distillation" produces better results than the "original" models, why don't the US model owners do it to themselves?
Have ChatGPT 5.6 talk to itself to produce 5.7 or whatever?
And since they already know which requests came from China or looked like distillation, can't they just replay those same prompts?
They already do and everyone has been using that concept for years now, it's known as RLAIF.
You can’t train on our models trained on human text, that is stealing the IP we stole first!
We demand that possible all avenues to observe, replicate, and resell exogenous value remain uninhibited until ROI and monopoly is achieved!
Companies that are hurt by open weight models fight against them. Companies that benefit from open weight models fight overregulation. Color me surprised.
Ok, great, so Microsoft believes in open weight models. So... uhh, why not release a few?
They did. Check out the Phi models (e.g., https://huggingface.co/microsoft/phi-4).
Great, how about some frontier models like China seems to be doing on a weekly basis at this point. Give us a CoPilot variant that's open source & open weights.
They could start by releasing the MAI-1 models they announced recently. https://microsoft.ai/models/
now open ur driver on linux
Some more discussion on NVIDIA crosspost: https://news.ycombinator.com/item?id=49035751
[dead]
"the companies with the most to lose saw which way the wind was blowing and switched sides" mightve been an apt title.
[flagged]
Ok, but please don't fulminate or post unsubstantive comments here.
https://news.ycombinator.com/newsguidelines.html
It's not worthless, you don't have to agree with P on everything, or make it personal - it works as strictly business - only where and when appropriate.
[flagged]
Regardless of the motivations behind the paper, the fact is that they are right. It's a topsy-turvy world where America closes up and China opens, and I hope it gets corrected soon.
And besides, it seems to me like the ethical choice. Given how these models are, in some very real sense, mechanical plagiators, built on the generosity of creators past and present, some of them now in danger of being replaced by the machine. I think the least the labs can do is open these models up. These and other such considerations were the reason OpenAI started with that name. Of course, it was questionable that those ideals would survive the encounter with generational wealth. Just look up what the founders of Google were saying about advertising when they were two students tinkering at an as of yet unproven tech. Same thing for OpenAI, self-interest speaks that much louder when there's real money on the table.
It just boggles the mind that people now make excuses for their all-too-predictable about-turn.
> I think the least the labs […]
Perhaps worth not calling them "labs". Are they not (for-profit) companies?
Of course but I think labs is a good term. Like a lot of terms like this it comes from history. The people working on AI at these places come predominantly from academia and predominantly do research. Of course now these companies are far more than just a “lab” and I can imagine their headcount may be more non-research folks at this point but most of the results are indeed coming from the “lab”-y areas
> It's a topsy-turvy world where America closes up and China opens
There's nothing "open" about China. Google, meta, openai, etc all blocked. Go visit and see how it goes when you try to access your gmail or open facebook. Try to use chatgpt. Try to get citizenship and see how that goes. China blocks many western companies with their great firewall and force internal similar products. This is smart, China wants to prioritize their own.
Your examples are the typical case, which is why OpenAI being closed and Kimi/DeepSeek/Qwen being open is topsy-turvy.
Im guessing parent meant more open in this particular case, i.e open weight models vs closed cloud hosted. Obviously china is not an open country.
He means specifically open vs closed AI model development, and how that’s topsy-turvy for the reasons you mention.
well let's be clear, there's nothing open about an open-weight model.
To me, "open these models up. " must mean provide all the data and supporting documentation required to reproduce the model. That would be "open". Postgres is open because you can download all the data and supporting documentation and reproduce the binary yourself. However, being able to only download a postgres binary would make it no longer open.
Additional training on top of an open-weight model sounds analogous to writing mods for minecraft. You may change some behavior but that doesn't make minecraft "open".
It's open, with a small 'o', and no moral judgements attached to it.
1. You have the weights, so you can run the model yourself on your own hardware. 2. You have the weights, so you can do post-training and shift those weights for your own purposes. It's not the same as training the model, but for many people its fine as what we want is a quantization, or a fine tune, or to create hybrid models.
What they don't give you are the training data, and reproduction instructions but... the toolchain to create the software has never been a part of 'Open Source'.
Even though it feels like a huge loophole, it's technically open source if you deliver the source code, without having a compiler that's available so you force them to recreate the toolchain from scratch.
To me, though, a better analogy is to research science where you'll be happy when they give you the full result set they compiled even if you don't get the raw data which may have IP or privacy concerns, or their often poorly documented lab notes so you can actually reproduce.
What you want to do is run your own experiment, and get your own results... not duplicate theirs directly. Even if reproduction is your aim in science, being unable to reproduce without copious notes sometimes points out that the original experimental process must have been flawed.
LLMs have the same issue. The creation process isn't entirely well documented, and the raw data can't be released since although the company have the right to use certain sources, they can't transfer those rights to others.
Weights are the cake and source is the recipe. Except you aren’t going to spend millions working through the recipe to arrive at the same exact cake anyway, so why do you care?
Sure, I'm all for opening the whole enchilada. But at least Chinese AI development is more open than the US one. And not just on the weights front, but AFAIK also more open when it comes to the techniques used and lessons learned in the process (DeepSeek at least is). Which is in a sense even more valuable and laudable.
The problem with that analogy is that compiling Postgres takes a few minutes. Training a model takes hundreds of thousands of dollars (at least) and specialized hardware.
Having a standardized training set is valuable, though.
Here here! Open weights is the term because it's close to open source but everyone thinks everyone else is an idiot and saying the model is downloadable but you can't recreate it is too difficult for our feeble minds to understand. There are actually open source models out there though, with datasets and training code.
I think it has more to do with the fact that there has to be some term to describe the concept of "model which has weights that are openly accessible"
Like, that's just a logical thing to call it. I don't believe anyone is making a judgement on the intelligence of the reader to call it "open weight" when it refers to weights that are openly available.
"Open source" would be a more appropriate term to describe a model which also includes the training source.
Indeed.
This is some bizarre victim inversion. The providers of closed models are the ones who are trying to use regulation to stop their open model competition, not the other way around.
This is how it is with all of these guys, their only principle is "what's good for me", and will twist all narratives to fit it
That's why competition is good though
I'm very confused by this comment, I don't know who you're referring to.
None of the people signed this have ever produced a frontier model at a given date (Which is to say its neither Google/OAI/Ant). The ones that sign are meta, musk (who is in the shovel selling business as well), hugging face obviously and few others
Note: Just pointing out the comment intent and nothing else
OK, I guess my confusion was that the effort to ban open weights models is what I would categorize as "groveling at the white house to try to stop their competition".
Google has produced Gemini Pro, Meta produced llama3-70B/llama3-405B and now has Muse, Musk has Grok. These are very capable models and to claim that they are not frontier is to stretch the meaning unless you want to say "#1 model" Gemini Pro at one point even if for a week, was #1 model.
I included Google along with Ant and OAI. OSS models have never been the highest performing model unless you refer to benchmaxxing.
Musk didn’t sign it.
nVidia did and they released good models – same with Meta, Microsoft, IBM, Mistral – all are signatories.
ha! who is groveling to the white house to stop competition? openai, anthropic and google, they are the losers. if they want to compete, compete.
I don't think that is what is happening.
Instead a bunch of tech companies are gathering to try to stop OpenAI and Anthropic fear-bouncing the White House and the Republican Congress into giving them regulatory capture and repeating the mistakes they are making around RISC-V.
Those mistakes won't just entrench two companies, they will entrench the bigger-better-faster-more model (closed companies making ever bigger cloud-bound models) when it is abundantly clear that enormous progress can still be made on smaller, even desktop-bound models (where, due to distribution, open weights are essentially inevitable).
Regulatory capture that stops open weights work will also have impacts on local and on-device AI work, as well as on academic research.
You can call out regulatory capture evwn when you would do the same if the shoe were on the other foot.
I thought the whole point of open software development was anyone could take your thing and improve it.
It would seem as if the community either isn't doing that or is relying on the Chinese to do that.
I thought this would be a push for a nationwide open weights effort / initiative, but I was surprised to see this:
> Distillation ... reflects a long tradition of learning from, building upon, and improving existing technologies, a tradition that has helped drive innovation since the rise of the open-source software movement. By contrast, unlawful efforts to extract value from closed models raise legitimate concerns. Those concerns should be addressed through targeted legal and commercial frameworks rather than sweeping restrictions on techniques that play an important role in AI innovation.
Sounds like they are saying "please protect our IP theft" that created closed weight frontier models in case we arbitrarily decide to close our models. But don't get rid of distillations in general so that we can all also keep benefiting from open models. We don't want to lose the ability to benefit from the work of others as we launder IP into closed models.
Surprised Linux Foundation kept their name on it with that.
"..a tradition that has helped drive innovation since the rise of the open-source software"
i've said this in other comments but the fact that these companies are trying to align with open-source when there's no "source" included with their models is pretty damning. I think it lays bare the absence of any kind of noble or righteous motive with respect to distillation.
Now that their lawyers have cornered everyone trying out open weight models and banned the ones they can, they put out a statement like this.
I agree with the message, but I'm not going to accept it from this particular crowd.
Microsoft, NVIDIA, Meta, Palantir, IBM...They have all been actively hostile to open source for decades, and have a history of embracing it only when convenient and profitable.
Microsoft benefited from its close partnership with OpenAI for years right up until it went sour. Where was this enthusiasm for open weights then?
Meta was developing open models and then abandoned that effort in search for profits. Muse Spark is now fully closed.
All these companies have the resouces to train and release frontier open weights models today, but choose not to. So spare me the marketing and virtue signaling.
I’ll never forgive Microsoft for their hostile actions against open source and open standards in the 90s. I’ll never believe any endorsement from them of open-anything.
Yeah, the endorsement logos at the bottom are majority bad actors with regards to openness in code. Many are bad actors with regards to openness in _society_.
Yeah, never ask a fox to count your chickens.