Open letters about AI development

Blog · Read

• Surveys three competing late-July open letters on AI development: “Open Weights and American AI Leadership” (Microsoft-led, 235 signers incl. NVIDIA, Amazon, OpenAI, July 24) defending open weights and distillation as safety-positive; Anthropic’s pointed non-signature and counter-statement warning about authoritarian misuse and “industrial-scale distillation operations”; and “Pacing the Frontier” (1,324 frontier-lab employee signers, July 28) calling for government-backed slowdowns on automated AI research.
• Notes Anthropic’s conspicuous absence from the first letter as itself a signal of the industry’s underlying split on openness vs. restraint.
• Cites supporting stats from the broader moment: Anthropic reports 80% of its code now written by Claude Code; OpenAI cut serving costs 20%; Kimi K3 was used to help design proprietary chip architecture.
• Takeaway: the AI industry doesn’t have one unified policy position — “safety” is being invoked to argue for opposite prescriptions (openness vs. slowdown) by different camps within the same industry.