Microsoft-led AI letter on open weights draws Anthropic pushback
On July 24th, 235 AI-adjacent companies, including NVIDIA, Amazon, Y Combinator, The Linux Foundation, and a later signer, OpenAI, put their names to an open letter titled 'Open Weights and American AI Leadership,' shepherded by Microsoft. The letter argues against restricting open-weight models on safety grounds. It says relying solely on closed models is not inherently safe: closed models can be breached, misused, or fail in ways outsiders cannot detect, and concentrating advanced capability behind a small number of closed models creates single points of failure, weakens competition, and leaves critical technology in a handful of providers' hands. Open weights, it argues, let a broad community of researchers and developers examine a model's behavior, find vulnerabilities, build safeguards, and improve it over time. The letter also defends distillation, the practice of training or improving one model using another model's outputs, calling it a widely used and legitimate technique for model improvement, evaluation, and validation that policymakers should not conflate with misappropriation.
Anthropic did not sign. Three days later it published its own response, 'Our position on open-weights models.' CEO Dario Amodei argued that authoritarian governments could build 'AI models that are more powerful than those built by the US' and that models could be 'misused to carry out cyberattacks or biological attacks.' He called for 'a crack down on industrial-scale distillation operations,' while also stating that 'Anthropic has never advocated for a ban on open-weights models.'
On July 28th a third letter, 'Pacing the Frontier,' was published with 1,324 signatures from employees of frontier AI companies, among them Jakub Pachocki (Chief Scientist, OpenAI), Ilya Sutskever (Safe Superintelligence Inc, previously OpenAI), and Anthropic's Dario Amodei and Jack Clark. Its core request: that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development. Simon Willison, who compiled this account from his own sponsors-only newsletter, writes that the signatories' concern is intense competitive pressure combined with AI progress accelerated by automated AI research itself. As evidence for why that risk is being taken more seriously, he points to Anthropic producing 80% of its own code with Claude Code, OpenAI's Sol reducing end-to-end serving costs by 20%, and Kimi K3 designing a chip to serve a nano model built on its own architecture.
Key facts
- 235 AI-adjacent companies, including NVIDIA, Amazon, Y Combinator and The Linux Foundation, signed 'Open Weights and American AI Leadership' on July 24th; OpenAI signed later.
- Anthropic did not sign; three days later CEO Dario Amodei published a response calling for a crackdown on industrial-scale distillation rather than a ban on open weights.
- A separate letter, 'Pacing the Frontier,' published July 28th, gathered 1,324 signatures from frontier AI employees, including OpenAI's Jakub Pachocki and Anthropic's Dario Amodei and Jack Clark, asking governments to help pace automated AI development.
- Cited as evidence of the pace of change: Anthropic produces 80% of its code with Claude Code, OpenAI's Sol cut serving costs by 20%, and Kimi K3 designed a chip serving a nano model built on its own architecture.
Why it matters
Three open letters inside two weeks expose a real fault line in how the AI industry wants governments to treat frontier development. Microsoft's coalition frames open weights as a safety asset: distributed scrutiny catches flaws a single closed lab might miss. Anthropic frames unrestricted distillation as the bigger risk, letting rivals or hostile states catch up fast by training on a frontier model's outputs. 'Pacing the Frontier,' backed by leaders at both OpenAI and Anthropic, adds a third position: slow the race itself rather than fight over openness.
Who it affects
Policymakers deciding whether to restrict open-weight releases or distillation face pressure from both directions at once: open-weight backers like NVIDIA, Amazon, Y Combinator and The Linux Foundation on one side, and closed-model leaders like Anthropic on the other. Developers who build on open weights, and researchers and companies that use distillation to train or evaluate models, would feel the effects of any resulting policy directly.
How to use it
No product ships from this story. The practical takeaway for founders and policy watchers is that any near-term US action on model openness or distillation will land somewhere between these three documents. Anyone tracking AI policy is best served reading the letters themselves, since the named signatories, from NVIDIA to Jakub Pachocki, show which arguments are gathering institutional backing.
How solid is it
The account draws on three named, dated public letters and an on-record response from Anthropic CEO Dario Amodei, compiled by Simon Willison from his own newsletter. The cited figures, 235 signatories, 1,324 signatures, an 80% share and a 20% cost cut, are stated plainly rather than tied to a published dataset, and the source does not explain what OpenAI's 'Sol' is or how Kimi K3's chip works.
Risks and caveats
The source names only five of the 235 signatories to the first letter and gives no exact calendar date for Anthropic's response beyond 'three days later.' It also does not say whether 'Pacing the Frontier' proposes any enforcement mechanism beyond asking government to support the development of pacing tools, so how any of this translates into actual policy remains open.
“Anthropic has never advocated for a ban on open-weights models”
— Dario Amodei, CEO of Anthropic, responding to 'Open Weights and American AI Leadership'