Anthropic、独自原則の島で孤立し関係にひび

Anthropic spent this winter and spring as the darling of America's AI boom, conquering the frontier while casting itself as the industry's moral conscience.

By summer, the Claude maker had found itself alone on an island of its own principles.

Why it matters: Anthropic is simultaneously the world's most valuable startup and its most isolated AI leader. The company's idiosyncrasies have strained its relationship with developers, policymakers and partners as it barrels toward a potential trillion-dollar IPO.

Driving the news: Anthropic is the only frontier AI lab that declined to sign an open letter led by Nvidia CEO Jensen Huang urging Washington not to restrict open-weight models, a category currently dominated by China.

Open-weight models are downloadable systems that users can run, modify and build on without paying for a company-controlled interface — making them cheaper and more flexible, but also harder to control once released.

Google and OpenAI, Anthropic's two biggest closed-model rivals, joined dozens of other signatories over the weekend, leaving Anthropic alone in defending the business model all three companies still depend on.

What they're saying: Anthropic CEO Dario Amodei tried to defuse the fight himself on Monday, publishing a blog post insisting Anthropic has never called for banning open models — while still not signing onto the letter.

Amodei called less capable open models "a public good" and said banning U.S. companies from using Chinese open models would do little to stop hostile governments or bad actors.

But he doubled down on the need for tighter controls on the ingredients and behaviors that could make powerful models dangerous, including advanced chips flowing to authoritarian governments.

Between the lines: The open-weight controversy captures the new bad blood roiling Silicon Valley, where style and posture matter just as much as policy.

Anthropic is more willing than its chief competitor, OpenAI, to be openly confrontational, even when the two companies are not always that far apart intellectually or commercially.

Zoom out: As the AI race accelerates, the company is growing isolated across nearly every major policy fight:

🪖 Military use: The Pentagon blacklisted Anthropic in February after a bitter fight over whether Claude could be used for mass surveillance or autonomous weapons. Top Pentagon official Emil Michael lashed out at Anthropic on Friday — claiming "there is no AI company more hostile to the warfighter."

⚠️ Guardrails: No lab has pushed harder for regulation than Anthropic, which has backed state-level AI laws and proposed an FAA-style regulator to screen advanced models before release. But when the White House flagged a security vulnerability this summer, Anthropic disputed how serious it was — triggering export controls that forced its Fable and Mythos models offline for nearly three weeks.

🧑‍🔬 Distillation: Anthropic has accused Chinese labs of training their lower-cost models on Claude's outputs via "industrial-scale distillation campaigns." Critics counter that distillation is a standard part of AI development — and that Anthropic is an awkward messenger after settling a $1.5 billion copyright lawsuit over pirated books.

The other side: Anthropic joined more than 1,100 AI employees from OpenAI, Google DeepMind, Meta and other rivals in a separate petition Tuesday urging the U.S. government to help develop tools to "deliberately pace" frontier AI development.

The letter shows that despite its isolation, Anthropic's top concern — that frontier AI is moving faster than industry or government can safely control — is increasingly shared across the labs racing to build it.

Reality check: For all the noise, Anthropic remains in an extraordinarily strong position — arguably the strongest of any AI lab in the world.

Its models still rank at the top of most independent benchmarks, and enterprise customers keep paying premium prices despite the backlash over cost and restrictions.

The company hit a $965 billion valuation in May, ahead of OpenAI, and is preparing for an IPO that could value it even higher.

For the record: In response to a request for comment, Anthropic pointed to Amodei's blog post defending its stance on the key issues.

The bottom line: The same qualities driving Anthropic's isolation — caution, refusal to bend, insistence on being right rather than liked — are the ones that helped establish its lead in the first place.