This winter and spring Anthropic emerged as one of the more celebrated startups in America's AI boom, advancing rapidly on frontier capabilities while positioning itself as a moral voice in the industry. By summer, however, the company increasingly found itself isolated because of its principles.
What happened
Anthropic was the only frontier AI lab that declined to sign an open letter led by Nvidia CEO Jensen Huang asking Washington not to restrict so-called "open-weight" models. Open-weight models are distributed as downloadable systems that users can run, modify and build on without paying for a company-controlled interface. They tend to be cheaper and more flexible but are harder to control once released.
Google and OpenAI, Anthropic's two largest closed-model competitors, joined dozens of other signatories over the weekend, leaving Anthropic as the sole major holdout.
Anthropic's position
CEO Dario Amodei published a blog post on Monday attempting to defuse the dispute, stressing that Anthropic has never advocated banning open models while still not signing the Nvidia-led letter. Amodei described less-capable open models as "a public good" and argued that banning U.S. firms from using Chinese open models would do little to stop hostile governments or bad actors.
At the same time he reiterated Anthropic's call for tighter controls over the components and behaviors that could make powerful models dangerous — including concerns about advanced chips reaching authoritarian states.
Political and security rifts
The open-weight debate highlights growing friction in Silicon Valley, where posture and public stance matter as much as policy. Anthropic has shown a greater willingness than some rivals, notably OpenAI, to take confrontational public positions even when the companies are not always far apart technically or commercially.
Anthropic has become isolated across multiple policy fronts in recent months:
-
Military use: The Pentagon blacklisted Anthropic in February after a contentious disagreement over whether Claude could be used for mass surveillance or autonomous weapons. Emil Michael, a senior Defense Department-affiliated official, publicly criticized the company.
-
Guardrails: Anthropic has been one of the most vocal proponents of regulation, supporting state-level AI laws and proposing an FAA-style regulator to screen advanced models before release. Yet when the White House flagged a security vulnerability this summer, Anthropic disputed its severity — a dispute that triggered export controls and forced its Fable and Mythos models offline for nearly three weeks.
-
Distillation: Anthropic accused Chinese labs of training lower-cost models on Claude outputs through "industrial-scale distillation campaigns." Critics note that distillation is a common practice in AI development and point to Anthropic's earlier settlement of a $1.5 billion copyright lawsuit related to pirated book content as complicating its claims.
Areas of agreement
Despite the tensions, Anthropic joined a separate petition supported by more than 1,100 AI employees from OpenAI, Google DeepMind, Meta and other rivals urging the U.S. government to help develop tools to "deliberately pace" frontier AI development. That joint letter indicates that Anthropic's central concern — that frontier AI is progressing faster than industry or government can safely manage — is shared across many leading labs.
Market position
Amid the noise, Anthropic remains in an exceptionally strong commercial and technical position. Its models continue to rank at or near the top in most independent benchmarks, and enterprise customers keep paying premium prices despite pushback over cost and restrictions. In May the company reached a $965 billion valuation, ahead of OpenAI, and it is preparing for an IPO that could value it even higher.
Conclusion
The same traits that have contributed to Anthropic's isolation — caution, a refusal to yield on principles and an insistence on being correct rather than popular — are also factors that helped build its lead. While the company remains powerful in technical and market terms, strained relations with developers, policymakers and partners could constrain its options going forward.
For the record, Anthropic pointed to Dario Amodei's blog post when asked for comment on the contested issues.
