Anthropic has told prospective shareholders that it may be close to a safety milestone it defined for itself. In Amendment No. 1 to its Form S-1 registration statement, filed with the US Securities and Exchange Commission on Monday, November 2, 2026, the company says its internal estimates and its expectations for future models, rather than its current ones, indicate that it may imminently meet the Responsible Scaling Policy's definition of "dramatically accelerated progress attributable to AI." It adds that it has not designated any current model as exceeding that threshold and is still assessing the underlying data.
The language sits in the risk factors. Anthropic writes that its growing use of AI to speed up its own research could push model capabilities ahead of its ability to develop, validate and deploy matching safety, alignment and security measures. Much of its evidence on that pace comes from lagging indicators, it says, so it has in the past failed to spot a significant acceleration until after the fact and may do so again. The company says it is evaluating what meeting its AI research and development capability threshold would imply and will publish further reports as the policy requires. TechCrunch, which first highlighted the passage, reports that these statements were added in the amendment.
Elsewhere in the business section, Anthropic describes Claude as one of the biggest drivers of its own progress. It says Claude now takes part in running experiments, generating training data, automating evaluations and safety audits, and writing and reviewing code for later model generations, a dynamic it calls closing the loop. The filing says more than 90% of contributions to its codebase in September 2026 were written by Claude, and that on its internal benchmarks Claude's success rate on hard, loosely specified engineering problems reached about 90% that month, roughly 50 percentage points higher than nine months earlier.
Anthropic also sets limits on that ambition in writing. It says it does not expect to be able to safely use Claude to design and build successive generations of frontier models fully on its own without meaningful human supervision or understanding, if that ever becomes technically possible, and that it is considering changes to its Responsible Scaling Policy to keep humans in control of frontier model development.
The amendment's main purpose was financial. The first prospectus, filed on October 26, covered results only through June. The new document reports revenue of $17.0 billion for the quarter ended September 30, up from $1.4 billion a year earlier, and an operating loss of $567 million after operating income of $974 million in the second quarter. Anthropic attributes the swing mainly to $5.0 billion of compute bought under short-term commitments, which it says costs much more than capacity under multi-year deals, and $2.0 billion of non-cash charges tied to a donation matching program. CNBC calculates that quarterly revenue grew about 47% from the prior quarter, down from 144% in the second quarter, and reports a third-quarter net loss of $23.29 billion.
The filing also puts a figure on Anthropic's long-term obligations: purchase commitments and lease agreements with an aggregate committed spend of about $895.4 billion, substantially all for compute and infrastructure. As of September 30 the company held $95.6 billion in cash and cash equivalents and $36.3 billion in short-term investments. The cover page still leaves the price range blank, and the company has applied to list on the Nasdaq Global Select Market under the ticker ANT.
The safety passage should be read for what it is. A registration statement is a legal disclosure document, and TechCrunch notes that such filings also serve a marketing purpose. Anthropic does not say that recursive self-improvement has started, and the threshold in question is one the company defines and assesses internally. What has changed is that a frontier developer about to go public now states, in a document carrying legal liability, that its own models may soon meet a bar its policy treats as calling for extra safeguards, and that it has missed such accelerations in the past.