Anthropic warned, in a document filed with the U.S. SEC on November 2, 2026, that it may be close to meeting its own internal threshold for sharply accelerated artificial intelligence research driven by AI itself. The warning appeared in the company's amended IPO prospectus, Form S-1/A, and was first reported by TechCrunch.
According to the text quoted in the press, Anthropic's internal estimates and expectations for future models — not current models — indicate that the company may imminently meet its Responsible Scaling Policy definition of “dramatically accelerated progress attributable to AI”. The company stresses that no current model has officially been designated as above that threshold and that it is still assessing the data.
What the threshold means
The threshold is not synonymous with a fully autonomous AI that rebuilds itself alone. Under the company's policy, it can be met when AI systems double the aggregate pace of capability progress through automated research and engineering, or replace the entire research workforce at competitive cost. Crossing it would therefore not necessarily mean achieving full recursive self-improvement.
For context, Anthropic said that in September 2026 Claude accounted for over 90% of contributions to the company's codebase, and that average daily code output per engineer in the third quarter was more than 45 times the 2024 level. These figures measure code volume, not automatically equivalent gains in engineering productivity.
What Anthropic already admits
The company also admits an uncomfortable point: in the past it failed to identify a significant acceleration in its development pace until after it had happened, and that could happen again. It therefore says it does not expect to safely use Claude to design and build successive generations of frontier models without meaningful human supervision, and is considering updating its policy to keep humans in control of development.
Editorial classification: the content of the SEC filing is CONFIRMED; the projections about future models are the company's own estimates, not an established fact.
Image credit: Anthropic, via Wccftech.
Transparency note: this article was prepared with the assistance of artificial intelligence, based on public sources and the internet sources listed below. We recommend consulting the original sources to confirm the details.
Muse
Fontes: Wccftech; Pondero; AI Affairs.
