THE ESSENTIALS IN 30 SECONDS

  • Anthropic's S-1/A, filed November 2, says internal expectations about future models suggest it may imminently meet its RSP definition of dramatically accelerated, AI-driven progress, according to TechCrunch.
  • The company says no current model has been designated as exceeding that threshold and it is still assessing the underlying data.
  • It acknowledges it has previously spotted big jumps in its development pace only after the fact, and says it is weighing RSP changes to keep humans in control.

Most of the attention on Anthropic's amended IPO registration, filed with the US Securities and Exchange Commission on November 2, went to its revenue and costs. But the updated risk factors also contain a notable safety disclosure: the company says it may soon reach the point its own policies associate with recursive self-improvement, where AI systems substantially speed up the development of better AI.

What the filing says

According to TechCrunch's reading of the S-1/A, Anthropic says it uses AI in its own research and development and warns that doing so could carry significant safety risk if it sharply accelerates progress. That risk is handled under its Responsible Scaling Policy (RSP), which sets out how the company monitors and mitigates dangers as capabilities grow.

The new language says Anthropic's internal estimates and expectations about future models, not current ones, indicate it may imminently meet the RSP definition of "dramatically accelerated progress attributable to AI." The company states it has not designated any existing model as exceeding that threshold and is still assessing the data behind its estimates.

The filing also concedes a monitoring weakness. Anthropic says that in the past it did not recognise significant accelerations in its pace of development until after they had happened, and that this could recur. It adds that it does not expect to be able to safely let Claude design and build successive frontier models fully autonomously without meaningful human oversight, and that it is considering RSP updates to keep humans in control.

How the policy frames it

Anthropic's RSP page lists version 3.4, effective July 8, 2026, which the company says revised its automated R&D threshold to better track the threat it worries about; the page does not spell out the new wording. Earlier versions described one AI R&D level as the ability to cause dramatic acceleration in effective scaling, later clarified as roughly doubling the rate of aggregate capability progress. For models past a lower AI R&D tier, the policy already demands an affirmative case explaining how misalignment risks are mitigated.

Context and caveats

The disclosure follows a June essay from the Anthropic Institute, reported by Sherwood News, that discussed scenarios in which AI largely builds its successors and said Claude was writing about 80% of the company's internal code. TechCrunch notes that skeptics see an S-1 as partly a marketing document, so claims of nearing self-building AI may also serve investors. The filing offers no public timeline or evidence that can be checked independently.