Anthropic Alone on Open-Weight Letter as Isolation Grows
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic was the only frontier AI lab to decline signing an open letter led by Nvidia CEO Jensen Huang urging Washington not to restrict open-weight models, a category currently dominated by China — while Google and OpenAI joined dozens of other signatories over the weekend.
- Dario Amodei published a blog post Monday insisting Anthropic has never called for banning open models, calling less capable open models "a public good," but doubled down on tighter controls on advanced chips flowing to authoritarian governments.
- The Pentagon blacklisted Anthropic in February after a fight over whether Claude could be used for mass surveillance or autonomous weapons; top Pentagon official Emil Michael claimed Friday there is "no AI company more hostile to the warfighter."
- A White House-flagged security vulnerability this summer triggered export controls that forced Anthropic's Fable and Mythos models offline for nearly three weeks after the company disputed how serious the issue was.
- Anthropic has accused Chinese labs of "industrial-scale distillation campaigns" training lower-cost models on Claude's outputs — a charge critics note is awkward given the company's $1.5 billion copyright lawsuit settlement over pirated books.
- Anthropic joined more than 1,100 AI employees from OpenAI, Google DeepMind, and Meta in a separate petition urging the U.S. government to help "deliberately pace" frontier AI development, showing its core concern about pace is increasingly shared across labs.
- Anthropic hit a $965 billion valuation in May, ahead of OpenAI, and is preparing for an IPO that could value it higher, with its models still topping most independent benchmarks.
Why it matters: Anthropic's refusal to sign the open-weight letter — a stance its CEO publicly softened without endorsing — leaves it alone defending the closed-model business model all three frontier labs depend on, while simultaneously battling the Pentagon, the White House, and Chinese distillation accusations. For enterprise customers and policymakers, the friction raises the practical cost of partnering with the best-performing model on most benchmarks.
