25 Tech Firms Including Nvidia, Microsoft, Meta Urge US Not to Restrict Open-Weight AI Models
- 25 companies including Nvidia, Microsoft, Meta, and Palantir signed a letter urging policymakers to avoid "premature restrictions" on open-weight AI models that could "stifle competition or drive innovation overseas".
- The letter comes as Chinese open-weight models gain ground, especially Moonshot AI's Kimi K3, which reportedly beats leading American models on some benchmarks.
- Treasury Secretary Scott Bessent said the Trump administration is investigating whether Chinese firms stole US intellectual property and that the government has the power to sanction them.
- OpenAI and Anthropic, both mainly closed-model companies preparing for possible massive IPOs, did not sign the letter.
- Nvidia CEO Jensen Huang, Microsoft CEO Satya Nadella, and Elon Musk all publicly amplified the letter on social media, though Musk's SpaceX did not officially sign it.
Hacker News 의견들
The real issue isn't open weights in general, it's Chinese models specifically, whether open or closed. Nothing in this article suggests otherwise. Hegseth himself said open weights matter if they're American made.
The debate is broader than that. OpenAI's head of strategic futures just said last week that open models are inherently decelerationist and ungovernable. Notice OpenAI isn't a signatory here despite its Microsoft ties.
Is the US planning to invade China and take their models away? Banning them in the US doesn't make China's models disappear, China isn't going along with the US position on this.
Why does anyone care what Hegseth thinks about open weight models? What's actually wrong with Chinese models, the censorship? I'd rather have diversity, US flavored output from US models and Chinese flavored from Chinese ones.
What even counts as a 'Chinese model'? If I change one weight in Kimi, fine tune it, distill it without transferring weights directly, or it gets released through a European company, is it still Chinese?
There's nothing open source about an open weight model honestly. It's like the binary blobs Linux fought against, if you can't train from source you're missing the actual 'source' part.
Nvidia talking about open source values is rich when they won't open source their Linux drivers. Feels like they're hijacking the open source narrative for something that has nothing to do with it.
I wonder what's really happening behind closed doors for these companies to put out a joint letter like this.
Different companies have different motivations here. A lot of them lost the race to OpenAI, Anthropic, and Google and know they can't compete, so they're shifting strategy.
If only two AI companies end up allowed to train or run frontier models, demand for hosting, GPUs, and data centers drops, so Nvidia benefits from more inference providers running Chinese open weight models instead.
My guess is they're worried about OpenAI and Anthropic building their own chips and controlling the whole stack.
Nvidia sells the tools so more models running anywhere is good for them. Microsoft and Meta are also-rans now, their best shot is leaning on their own infra and open models.
Most companies will probably end up running local models eventually, otherwise everyone becomes totally dependent on Anthropic and that dependence gets milked once it's entrenched.
This looks like a real push against sanctions or a commerce block on Chinese models. Anthropic just doubled its political spending to $40 million for the midterms to push 'AI safety' framing.
These companies see officials parroting OpenAI and Anthropic's lobbying talking points and they can smell regulatory capture coming.