Anthropic remains alone in every AI battle, yet emerges victorious regardless.

      Six months ago, Anthropic was regarded as the star of America’s AI surge, leading in model development and taking a strong stance on ethical issues. This summer, however, it finds itself in a position of isolation.

      According to Axios, Anthropic, the maker of Claude, is now the outlier in nearly every major AI policy debate. While it holds the title of the world's most valuable startup, it also stands as the most isolated laboratory in the sector, with these two aspects being interconnected.

      The most evident sign of this trend emerged last week when Anthropic was the sole leading lab that chose not to endorse an open letter encouraging the U.S. government to avoid imposing restrictions on open-weight models. Nvidia’s Jensen Huang spearheaded the effort, and both Google and OpenAI, its closest competitors, added their signatures. Consequently, Anthropic found itself advocating for stricter regulations on the very business model all three companies still rely upon.

      Dario Amodei attempted to soothe the controversy personally. In a blog post, Anthropic's CEO asserted that the company has never called for a ban on open models, even labeling less robust models as “a public good.” Nonetheless, he still refrained from signing the letter and maintained that tighter controls were necessary for potentially dangerous elements, particularly advanced chips being acquired by authoritarian regimes.

      The pattern of isolation continues on other fronts as well. In February, the Pentagon placed Anthropic on a blacklist following a dispute over whether Claude could be utilized for surveillance or autonomous weapons, with a senior defense official stating last Friday that "no AI company is more hostile to the warfighter."

      On safety matters, Anthropic has been more proactive than any other entity, endorsing state AI regulations and proposing a regulatory body similar to the FAA to review models before they are released. However, when the White House flagged a security vulnerability this summer, Anthropic contested the severity of the issue, which ultimately contributed to the implementation of export controls that took two of its models offline for nearly three weeks.

      Additionally, there are concerns regarding distillation. Anthropic accused Chinese laboratories of conducting “industrial-scale” campaigns to train cheaper models using outputs from Claude, as reported by the New York Times. Critics argue that distillation is a normal aspect of AI development and point out the awkward timing of these accusations—Anthropic had just resolved a $1.5 billion copyright lawsuit concerning pirated books that were used to train its own models.

      Despite its isolation, Anthropic is not completely alone. This week, it joined over 1,100 employees from competing labs in a petition urging the government to assist in "deliberately pacing" frontier AI. Its primary concern, that the technology is advancing more rapidly than can be safely managed, is now a common sentiment within the industry.

      Despite the tensions, Anthropic is not faltering. Its models continue to excel in most independent assessments, and businesses continue to pay premium prices, despite dissatisfaction surrounding costs and regulations. In May, it reached a valuation of $965 billion, surpassing OpenAI, and is on track for an IPO that may raise its valuation even further.

      The traits that contribute to its isolation in policy debates—caution and an unwillingness to compromise—are precisely the qualities that have propelled its success. Anthropic prefers to be correct rather than popular, and for the time being, it can afford that stance.