Key points
- Anthropic CEO Dario Amodei published a post on July 27 saying the company "has never advocated for a ban on open-weights models."
- He backs three measures instead: no powerful chips sold to China, a crackdown on industrial-scale distillation, and mandatory safety testing for every sufficiently capable model, open or closed.
- He rejects two claims in the Nvidia-hosted industry letter, that open weights make safeguards easier and that broad access helps defenders more than attackers.
- That letter has grown from 25 signatories to 133 since July 24. Amazon (AMZN) has signed. Anthropic and xAI have not.
Anthropic chief executive Dario Amodei published a post on Monday, July 27, titled "Our position on open-weights models."
"Anthropic has never advocated for a ban on open-weights models," he wrote.
It follows a letter published three days earlier. Nvidia (NVDA) chief executive Jensen Huang shared the letter, titled "Open Weights and American AI Leadership," on July 24, in his first post on X. Twenty-five companies signed it that morning. OpenAI added its name after the launch. Anthropic didn't sign and declined to comment at the time.
Amodei wrote that "some people have even accused Anthropic of wanting to ban open-weights models as a means of protecting our business." On that point he wrote that a ban "would protect US AI companies from competition, but that has never been my goal."
On the models themselves: "Open-weights models that don't have dangerous capabilities are a public good: they don't cost anything besides the compute needed to run them, and they provide value to businesses, developers, and researchers."
The two risks he named
Amodei described two concerns, both of which he said he laid out in his essay The Adolescence of Technology six months ago.
The first is that authoritarian governments build more powerful models than the US does. "It is irrelevant whether these models are released with open weights, and certainly irrelevant whether they are used by US businesses," he wrote. "In fact, the most dangerous model may be one that is trained in secret and handed only to the People's Liberation Army for use in drones and the Ministry of State Security for surveillance and repression."
He cited two US government sources for that view. Vice President JD Vance said in Paris in 2025 that "authoritarian regimes have stolen and used AI to strengthen their military, intelligence, and surveillance capabilities." The Intelligence Community's 2026 Annual Threat Assessment found that "other global powers' robust progress in AI is challenging US economic competitiveness and national security advantages."
The second concern is cyberattacks, biological attacks and alignment failures. Amodei granted that open models can carry more of that risk, because guardrails are hard to apply and released weights can't be pulled back. Then he ruled out the remedy under discussion in Washington. "Banning the use of these models by US businesses does nothing to address this risk, because bad actors are unlikely to be legitimate US businesses."
The three measures he supports
| Measure | Amodei's stated reason |
|---|---|
| No sales of powerful chips or chipmaking equipment to China, plus a crackdown on smuggling | China "cannot build more powerful models than the US without US chips" |
| Policy action against industrial-scale distillation operations | Distillation "can bring the Chinese frontier to within a few months of the US frontier" |
| Mandatory safety testing of all sufficiently capable models, open and closed | Test "for cyber, biological, and alignment risks before release" |
Distillation is the practice of training one model on another model's outputs. Amodei wrote that it lets China "build much better models than its number of chips would ordinarily enable, and thus partially evade chip bans."
On testing, he said the idea is close to consensus, that the Trump administration has moved toward it, and that it would have to be global. "Which means even the CCP would need to be on board," he wrote. He pointed to recent Anthropic research on modular training strategies as one route to making open-weight models safer.
Nvidia is the largest US seller of AI chips. Huang told Axios on July 22 that Chinese open-weight models such as Moonshot AI's Kimi are "excellent" and that American companies should "absolutely" be allowed to use them.
Where he disagrees with the letter
Amodei said he agrees with much of the letter, including that open weights expand access to the AI economy and give customers more control. Two claims he rejected.
"I don't agree with the letter's assertions that open-weights models necessarily make it easier to develop safeguards or that broad access to capabilities necessarily helps defenders more than attackers," he wrote. "It seems at least as likely to me that the opposite will be true."
His example was biology. "I worry that biology will have a strong attacker-defender asymmetry, where sufficiently capable models may be able to quickly weaponize pandemic-level viruses with widely available materials, whereas defense against these agents is a multi-year operational task in the best case," he wrote, citing Operation Warp Speed.
The letter calls distillation "a widely used technique for model improvement, evaluation, and validation." Amodei's phrase for the version he wants stopped is "industrial-scale distillation operations." He added: "A blanket ban on open-weights models is neither the correct remedy nor something we have called for."
An Anthropic researcher went after the signers first
Julian Schrittwieser, a member of technical staff at Anthropic, replied to the letter posts on X on July 24, before his CEO said anything. He told Microsoft (MSFT) chief executive Satya Nadella: "And @satyanadella coming out for open source too, can't wait for the open sourcing of Windows and MS Office!" He made the same joke at Nvidia about a CUDA and GPU driver open source release.
He then clarified what he meant.
funny to see people jump to the conclusion I must want to ban open weight models.. I actually think open models can be very useful! But it's interesting how some historically extremely anti-open source companies are suddenly all in favor of openness
See Julian Schrittwieser's post on X
The letter is now at 133 signatories
The list has grown more than fivefold in three days. It carried 25 names at launch on July 24. Forbes contributor Sandy Carter counted 50 on July 25. The copy hosted by Microsoft (MSFT) now lists 133.
Amazon (AMZN) is on it. So is Google, a unit of Alphabet (GOOGL), whose chief executive Sundar Pichai backed the letter on July 25 before the company itself appeared. Other names added after the launch include AMD, Intel (INTC) and SpaceX. Cisco (CSCO), Cloudflare (NET) and Palo Alto Networks (PANW) signed as well. The list runs past AI and technology names, into SAP, Siemens, Comcast (CMCSA) and Uber (UBER).
Anthropic is not on the list. Neither is xAI.
What a ban would touch
Amodei's own summary of the trigger: "Reports suggest that some US officials are considering banning the use of Chinese open-weights models by US companies."
Microsoft has been testing Moonshot's Kimi K3 for Copilot, a move it valued at about $600 million a year off its Azure AI bill. Jim Cramer said this month that Chinese AI models are run by the PLA and that US companies must stop using them. Treasury Secretary Scott Bessent has floated sanctions against overseas AI developers.
How the stocks closed
| Company | July 24 close | July 27 close | Change |
|---|---|---|---|
| Nvidia (NVDA) | $206.84 | $196.51 | -5.0% |
| AMD | $521.95 | $494.95 | -5.2% |
| Microsoft (MSFT) | $381.70 | $389.10 | +1.9% |
| Meta (META) | $595.19 | $593.87 | -0.2% |
| Alphabet (GOOGL) | $319.74 | $326.56 | +2.1% |
Monday's chip decline traced to other news. Three separate headlines hit semiconductors at the open: a report on Chinese lithography machines, CXMT's Shanghai listing, and a report that Nvidia may guarantee $250 billion of OpenAI's data center financing.
Microsoft and Meta report earnings after the close on Wednesday, July 29. Nvidia reports on August 26.
Sources
- Dario Amodei, "Our position on open-weights models," Anthropic, July 27, 2026
- "Open Weights and American AI Leadership," dated July 24, 2026, and its published signatory list
- Julian Schrittwieser's posts on X, July 24, 2026
- Forbes, "Huang's Open Weights Letter Doubled To 50 Without Amazon And Anthropic," July 25, 2026
- Closing prices from Yahoo Finance
This is general market commentary and opinion, not investment advice. Markets can go down as well as up, and you can lose money. Always do your own research and consider speaking with a licensed financial professional before making any investment decision.
