Anthropic stands alone in every AI battle, yet still manages to succeed.

      Six months ago, Anthropic was at the forefront of America's AI boom, excelling in model development and vocal about ethical considerations. However, this summer, the company finds itself increasingly isolated.

      According to Axios, the creator of Claude is now the outlier in almost every significant AI policy debate. It holds the title of the world's most valuable startup while simultaneously being the most alone in the industry. These two situations are evidently linked.

      The most notable indication came last week when Anthropic was the sole leading lab that chose not to sign an open letter urging Washington to avoid placing restrictions on open-weight models, which was spearheaded by Nvidia's Jensen Huang. Both Google and OpenAI, Anthropic's closest competitors, signed the letter, leaving Anthropic in the position of advocating for stricter regulations on the very business model that all three companies rely on.

      Dario Amodei, the CEO of Anthropic, attempted to mitigate the controversy himself. In a blog post, he asserted that the company has never advocated for a ban on open models and referred to weaker models as "a public good." Nonetheless, he still refrained from signing the letter and maintained his position that there should be tighter control on hazardous components, particularly advanced chips that reach authoritarian regimes.

      The trend of isolation continues in other areas as well. In February, the Pentagon blacklisted Anthropic following disagreements over whether Claude could be utilized for surveillance or autonomous weapons. A senior defense official remarked on Friday that there is “no AI company more hostile to the warfighter.”

      On the safety front, Anthropic has been the most proactive, supporting state AI legislation and proposing an FAA-style regulatory body to evaluate models before their launch. However, when the White House identified a security vulnerability this summer, Anthropic contested the severity of the issue. This disagreement contributed to the implementation of export controls that temporarily took two of its models offline for nearly three weeks.

      Moreover, Anthropic has accused Chinese labs of training cheaper models using Claude’s outputs in “industrial-scale” campaigns, as reported by the New York Times. Critics argue that distillation is a standard aspect of AI development and highlight the unfortunate timing, given that Anthropic had just reached a $1.5 billion settlement regarding copyright infringement over pirated books utilized to train its own models.

      Nevertheless, Anthropic's isolation isn't absolute. It joined over 1,100 employees from competing labs this week in a petition urging Washington to help “deliberately pace” the advancement of frontier AI. The primary concern, that the technology is progressing faster than can be safely managed, is now a shared sentiment within the industry.

      Despite the tensions, Anthropic is not suffering losses. Its models consistently rank highly in independent evaluations, and businesses continue to pay premium prices despite complaints regarding costs and restrictions. It achieved a valuation of $965 billion in May, surpassing OpenAI, and is moving towards an IPO that could potentially elevate its value even further.

      The characteristics that contribute to its isolation in policy debates—caution and an unwillingness to compromise—are also the same traits that have allowed it to establish its leadership position. Anthropic prefers to be correct rather than popular, and for now, it is in a position where it can afford to maintain this stance.