Anthropic has warned that artificial intelligence could soon begin accelerating its own development at a pace that may outstrip the ability of researchers to keep up with safety measures.

The warning appeared in an updated S-1A filing submitted as part of the AI company’s preparations for a potential public listing. While the filing attracted attention for its financial disclosures, its comments on AI-driven research and development raised a separate concern about the future of advanced AI.

According to the filing, Anthropic’s increasing use of AI systems in its own research could cause model capabilities to advance faster than the company can develop, test and implement corresponding safety, alignment and security measures.

The company also acknowledged that it had previously failed to recognise significant increases in its development pace until after they had occurred.

More notably, Anthropic said its internal estimates and expectations for future models suggest it may soon reach the threshold in its Responsible Scaling Policy for what it calls a dramatic acceleration in AI-attributable capability progress.

This concept is closely associated with recursive self-improvement, a scenario in which AI systems become increasingly capable of helping researchers develop more advanced AI systems, potentially creating a feedback loop that speeds up technological progress.

However, Anthropic clarified that none of its current models has been designated as having crossed that threshold. The company is still assessing the underlying data, and its warning is based on expectations about future models rather than a claim that fully self-improving AI already exists.

Anthropic also said it does not expect to safely use Claude to design and build successive generations of frontier AI models entirely on its own without meaningful human supervision or understanding. It is considering updates to its safety policy to ensure people remain involved in the development of increasingly capable systems.

The disclosure adds to growing industry discussions about whether AI development is moving faster than the safeguards designed to manage its risks. While AI-assisted research could accelerate scientific discoveries and improve productivity, it also raises questions about human oversight and the ability to identify risks before systems become significantly more capable.

For Anthropic, the challenge will be to continue advancing its technology while ensuring that safety measures develop at a pace that can keep up.