On November 2, 2026, Anthropic, PBC filed an amended Form S-1 (S-1/A) with the US Securities and Exchange Commission. The original public filing of October 26 covered financials only through June; this site's October 27 entry covered its list of observed misaligned behaviours. The amendment adds the third quarter and, according to TechCrunch, new language in the Risk Factors section about using AI to build AI. We could not retrieve the filing's full text directly, so the passages below are quoted as TechCrunch reproduced them, and the figures as CNBC reported them, with the SEC link listed for readers who want to check.
The new risk-factor language
The filing says Anthropic uses AI in its own research and development and warns: "Using AI to help build more capable AI could also pose a considerable safety risk, particularly if it sharply accelerates the pace of progress." The passage TechCrunch highlights continues: "our increasing use of AI systems to accelerate our own research and development may cause the capabilities of our models to advance more rapidly than our ability to develop, validate, and implement corresponding safety, alignment, and security measures. As a result, we have in the past not identified, and may in the future not identify, a significant acceleration in our rate of development until after it has occurred."
It then gives an example: "our internal estimates and our expectations regarding future models, rather than our current models, indicate that we may imminently meet our Responsible Scaling Policy's definition of 'dramatically accelerated progress attributable to AI.'" The next sentence limits the claim: "We have not designated any of our current models as exceeding this threshold, and we continue to assess the underlying data." Finally: "We do not anticipate being able to safely use Claude to fully autonomously design and build successive generations of frontier models without meaningful human supervision or understanding, if and when that becomes technically possible. Accordingly, we are considering updates to our Responsible Scaling Policy to keep humans in control of frontier model development at Anthropic."
The third-quarter figures
CNBC reports that the amendment shows "$33.24 billion in revenue for the nine months ended Sept. 30" and "roughly $17 billion in revenue" in the third quarter, "up about 1,127%" from a year earlier. Quarter-on-quarter growth slowed: "Third-quarter revenue increased around 47%, a steep drop from 144% growth in Q2 and 132% in Q1." Anthropic reported a third-quarter net loss of $23.29 billion. It has "entered into purchase commitments and lease agreements with an aggregate committed spend of about $895.4 billion, 'substantially all of which is related to compute and infrastructure.'" Compute and infrastructure expenses rose "by $14.4 billion for the nine months ended Sept. 30, up 329%".
The following is our arithmetic from the reported figures. If the third quarter's roughly $17 billion were repeated for four quarters, annual revenue would be about $68 billion. The $895.4 billion in commitments is about 13 times that figure. Committing to compute on that scale only makes sense if the company expects its models to keep getting much better, and that expectation is the same one the risk factor describes as a safety concern.
What remains uncertain. The filing does not say which internal estimates point towards the threshold, what data Anthropic is still assessing, or when it would decide. A prospectus is a legal document written partly to limit liability and partly to sell shares, and TechCrunch notes that the RSI language could be read as marketing as well as disclosure. The Responsible Scaling Policy changes are described as under consideration, not adopted. The commitments span years, and the filing summary we read does not give the schedule.
Analysis: The sentence that teams can act on is the admission about timing: an acceleration may only be visible "after it has occurred". The same problem applies on a much smaller scale to anyone whose systems depend on a model that is improving quickly. At Belgavi's Book Store, Geetha's team reviews the help-desk agent each quarter, and could instead track a few leading signals every week: the share of CASE tickets the agent closes without escalation, the share of its refunds that a person reverses, and how often it touches tools it was not expected to need. If any of these moves sharply between two model updates, they should review the change before the next update rather than wait for the quarterly review. Waiting for a scheduled review is the failure the filing describes: noticing the jump only after it happened.
Sources and evidence
- Anthropic, PBC Form S-1/A amended registration statement (SEC EDGAR, filed November 2)
- Anthropic warns that it's very close to RSI capability, aka self-building AI (TechCrunch, November 2)
- Anthropic reports $17 billion in Q3 revenue, purchase commitments of $895.4 billion in revised S-1 (CNBC, November 2)
Anthropic's amended S-1 of November 2, as quoted by TechCrunch, warns that AI-accelerated R&D may outpace its safety work, that it has in the past noticed significant accelerations only after they happened, and that estimates about future models suggest it may imminently meet its Responsible Scaling Policy's definition of dramatically accelerated progress attributable to AI. No current model is designated as past that threshold. CNBC reports roughly $17 billion of third-quarter revenue and about $895.4 billion of commitments, substantially all for compute and infrastructure.