Anthropic's IPO Filing Warns It May Be 'Imminently' Close to Recursive Self-Improvement — But Its Own Data Shows the Loop Isn't Closed

Anthropic’s revised S-1A filing with the SEC warns the company may “imminently” meet its internal definition of recursive self-improvement, though it says no current model has crossed that threshold and it is weighing updates to its Responsible Scaling Policy. The disclosure comes as Anthropic moves toward a Nasdaq listing under the symbol ANT, after filing its initial S-1 on October 26; the Financial Times reported in August that it aimed to list in October at a valuation around $2 trillion. Anthropic’s own May 2026 essay, “When AI builds itself,” is more measured, stating “We are not there yet, and recursive self-improvement is not inevitable,” and distinguishing AI that speeds research from AI that decides what to research. Its published data show Claude leads 26% of AI research tasks as of August 2026, up from under 1% in February, while fully autonomous work remains zero in every area measured. In a weak-to-strong supervision test, agents closed 97% of the performance gap versus 23% by two human researchers in a week, but the result did not transfer cleanly to production-scale models and humans chose the problem and scoring rubric. More than 80% of merged code lines are now authored by Claude, and the filing notes Anthropic has “in the past not identified” a significant acceleration in its development rate until after it occurred.