Anthropic's lonely island

Anthropic is the only major AI lab that refused to sign an open letter this week opposing restrictions on open-weight models—leaving it at odds with Google, OpenAI, and dozens of other companies. Open-weight models are downloadable and modifiable AI systems anyone can run and build on; they're cheaper and more flexible than proprietary alternatives, but much harder to control once released. Per Axios, CEO Dario Amodei defended the refusal in a blog post Monday, insisting his company never called for banning open models but doubling down on the need for tighter controls on advanced chips flowing to authoritarian governments. This week's conflict captures a broader pattern: Anthropic has found itself isolated across nearly every major AI policy battle. The Pentagon blacklisted the company in February over concerns about surveillance and autonomous weapons. It pushed hard for regulation but then disputed a White House security vulnerability, triggering export controls that forced its Fable and Mythos models offline for nearly three weeks. Despite the friction, Anthropic remains extraordinarily strong: its models top independent benchmarks, enterprise customers keep paying premium prices, and the company hit a nine-hundred-sixty-five billion dollar valuation in May—ahead of OpenAI. It's preparing for an IPO that could value it even higher. The same qualities driving Anthropic's isolation—caution, refusal to bend, insistence on being right rather than liked—also helped establish its lead in the first place.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton