235 companies back open weights as Anthropic calls for a distillation crackdown
Microsoft's July 24 letter 'Open Weights and American AI Leadership' drew 235 signatures — NVIDIA, Amazon, Y Combinator, the Linux Foundation, and a late-signing OpenAI — arguing that open weights let a broad community examine model behavior, find vulnerabilities, and build safeguards. Anthropic declined to sign and answered three days later, with Dario Amodei citing authoritarian-government and cyberattack risk and asking for a crackdown on industrial-scale distillation operations, while maintaining Anthropic has never sought an open-weights ban. On July 28, 1,324 frontier-lab employees — including Jakub Pachocki, Ilya Sutskever, and Amodei — signed 'Pacing the Frontier,' asking the US government to support an international effort to deliberately pace automated AI development. The policy fault line now runs through the labs rather than between industry and regulators: OpenAI backs open weights in one letter while its chief scientist asks for pacing in another.
Source: simonwillison.net ↗
Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect.
Why this matters
- → Open weights debate splits frontier labs internally, not just industry vs. regulators
- → Distillation crackdown could restrict how AI companies improve models through training
- → Pacing proposal signals researcher concern about competitive pressure accelerating unsafe development