$965 billion. That is the May valuation Anthropic carried into summer as it held the frontier benchmark lead and found itself the only major AI lab absent from an open-weight policy letter co-led by Nvidia CEO Jensen Huang. The company is preparing for an IPO that could push that figure past $1 trillion, yet its refusal to join Google and OpenAI on the open-weight question has widened rifts with developers, policymakers, and partners at the same moment it is strongest commercially.
The open-weight letter Anthropic skipped
Google and OpenAI, Anthropic's two largest closed-model rivals, signed the letter urging Washington not to restrict open-weight models. Dozens of other companies joined. Anthropic declined. CEO Dario Amodei addressed the gap Monday in a blog post, saying the company has never called for banning open models and describing less capable open models as "a public good." He did not sign.
Amodei argued that banning U.S. companies from using Chinese open-weight models would not prevent hostile governments or bad actors from obtaining them. He held firm on the need for export controls on advanced chips flowing to authoritarian governments.
Policy disputes stacking up
The open-weight split is the newest friction point in a longer list. The Pentagon blacklisted Anthropic in February after a dispute over whether Claude could be used for mass surveillance or autonomous weapons. Pentagon official Emil Michael said Friday that "there is no AI company more hostile to the warfighter."
This summer brought another collision. After the White House flagged a security vulnerability, Anthropic disputed the severity. That disagreement triggered export controls that took Anthropic's Fable and Mythos models offline for nearly three weeks.
A third front involves training data. Anthropic has accused Chinese labs of running "industrial-scale distillation campaigns" using Claude's outputs. Critics noted the awkward optics: Anthropic settled a $1.5 billion copyright lawsuit over pirated books used in its own training.
Where the isolation ends
More than 1,100 AI employees from OpenAI, Google DeepMind, Meta, and other labs joined Anthropic signatories on a separate petition Tuesday, asking the U.S. government to develop tools to "deliberately pace" frontier AI development. Enterprise customers continue paying premium prices despite the policy noise, and benchmark rankings remain near the top of independent evaluations. The $965 billion valuation, established in May, is the number Anthropic is carrying into its IPO preparation.