Three open letters in two weeks pulled back the curtain on a real fault line inside the AI industry, and the biggest names picked opposite sides. According to Simon Willison, who summarized the exchange for readers, the fight is about open weight models: whether the US should welcome them, restrict them, or slow the whole race down. What stands out here is who signed which letter, and who pointedly stayed away.
Round one: the pro-open camp
The first letter, “Open Weights and American AI Leadership,” was shepherded by Microsoft, dated July 24th, and signed by 235 AI-adjacent companies. The list is heavy: NVIDIA, Amazon, Y Combinator, The Linux Foundation, and OpenAI as a later signer. Simon Willison reads it as a direct push against any instinct in Washington to ban or limit open weight models over safety worries.
The core argument is that closed models aren’t automatically safer. As the letter puts it, they “can be breached, misused, or fail in ways that outsiders cannot detect.” Concentrating power behind a few closed models, the signers say, creates single points of failure and weakens competition. Open weights let a broad community inspect behavior, find vulnerabilities, and build safeguards.
One surprise, per Willison: the letter defends distillation, the practice of training one model on another model’s outputs. The signers warn policymakers not to “conflate legitimate model-development techniques with misappropriation,” framing distillation as a normal step in building on existing technology.
Round two: Anthropic pushes back
Anthropic’s name is missing from that list. Three days later the company published its own response, “Our position on open-weights models.”
CEO Dario Amodei took a harder line. He stressed the risk of authoritarian governments building “AI models that are more powerful than those built by the US,” and models being “misused to carry out cyberattacks or biological attacks.” He called for “a crack down on industrial-scale distillation operations.” Notably, he also stated that “Anthropic has never advocated for a ban on open-weights models.”
So the disagreement is narrower than it looks. Both camps claim to support open weights. They split on distillation and on how much the government should police it.
Round three: slow down the frontier
Then on July 28th came “Pacing the Frontier,” signed by what the letter describes as “1,324 employees of frontier AI companies.” The names cross company lines: Jakub Pachocki (Chief Scientist, OpenAI), Ilya Sutskever (Safe Superintelligence Inc), Dario Amodei and Jack Clark (Anthropic), and more.
Their ask is blunt. They want the US government to “support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.” The worry, Willison notes, is competitive pressure colliding with AI that increasingly speeds up its own research. When labs report large chunks of their code and cost savings now coming from their own models, that self-accelerating loop stops being hypothetical.
Why this matters now
This is significant because the letters mark the moment the industry stopped speaking with one voice on regulation. The old story was “labs vs. regulators.” The new story is labs disagreeing with each other, in public, in front of policymakers who are actively writing the rules.
A few practical takeaways:
- If you build on open weights, watch the distillation debate closely. That’s where restrictions are most likely to land first, and it directly affects fine-tuning and model-improvement workflows.
- If you run a business on closed APIs, don’t treat that as the safe default. The Microsoft-led letter argues concentration is its own risk. Diversifying across providers is a hedge worth pricing in.
- If you’re tracking policy, note that Anthropic is positioning itself as the safety-first outlier while nearly everyone else pushes openness. Regulators tend to reward whoever frames the risk story first.
My read: the “pace the frontier” letter is the one to watch. When rivals sign the same page asking to be slowed down, they’re admitting the race has a speed none of them fully controls.
For the full breakdown of all three letters, see the original write-up from Simon Willison.