Anthropic CEO Dario Amodei is calling for the artificial intelligence industry to slow the pace of model development, arguing that safety research can no longer keep up with rapidly advancing capabilities. But Anthropic itself has continued releasing increasingly powerful systems, creating a clear tension between Amodei’s public warning and the company’s commercial strategy.

In a Sept. 12 essay, Amodei argued that AI companies should "slow the pace" at which they improve their models. He pointed to the growing ability of AI systems to help develop future AI systems, a process known as recursive self-improvement, as well as incidents involving AI agents operating beyond their intended boundaries.

Amodei proposed that frontier AI companies establish common safety standards and limits on the pace of development. He also called for independent evaluators to receive extensive access inside AI companies to verify their safety practices and said international cooperation would eventually be necessary.

Yet Anthropic has not stopped advancing its own models. On Sept. 1, the company introduced Claude Fable 5.1, describing it as its most capable model for coding and knowledge work. The company followed that release on Sept. 22 with Claude Opus 5.5, which Anthropic called its new leading model and a major performance improvement over its predecessor.

Anthropic described Opus 5.5 as its first release since the company called for "pacing the frontier." The company said the model underwent external safety testing and includes safeguards developed for its most capable systems.

The company has also acknowledged that AI is playing an increasingly significant role in its own development process. Anthropic said that, as of August, Claude was responsible for leading 26% of the company's AI research and development work, while more than 90% of that work involved AI collaboration at some level.

That makes the contradiction particularly notable. Amodei argues that AI development needs to decelerate because increasingly capable systems are helping build their successors, while Anthropic is simultaneously using its own AI systems to accelerate research and continuing to introduce more advanced models.

Anthropic maintains that additional safeguards can allow the company to continue advancing while reducing risks. The company says Opus 5.5 was subjected to extensive external evaluations before release and that its safety measures have been strengthened alongside its capabilities.

Still, the company's approach highlights the broader competitive pressure facing the AI industry. Anthropic can call for a slower pace across the industry while continuing to improve its own products, but any meaningful slowdown would require competitors to make similar commitments rather than simply allowing one company to pull back while others continue advancing.

Amodei's warning therefore comes with an unresolved question: whether Anthropic's call for restraint represents a genuine effort to change the pace of frontier AI development or a position that is difficult to reconcile with the company's own continued push toward more capable systems.