Anthropic’s first IPO filing said it might not notice a jump in its own AI’s progress until after it happened. The amended version says that’s already happened.
That line sits in the amended S-1 Anthropic sent the SEC on November 2, next to a new warning that the company “may imminently meet” its own definition of “dramatically accelerated progress attributable to AI.” That’s Anthropic’s tripwire for AI that speeds up building the next AI.
The filing names no model and no date, and Anthropic plans to list on Nasdaq under the ticker ANT.
The New Risk Factor
The warning is part of the very first risk factor in the prospectus, the one about keeping Anthropic’s models at the front of the pack. Here’s the new part, word for word:
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. For 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.” We have not designated any of our current models as exceeding this threshold, and we continue to assess the underlying data.
Anthropic adds that it’s “evaluating the implications of meeting the AI research and development capability threshold” and will publish more reports under its policy.
What Changed Since October
The October 26 filing already warned that much of Anthropic’s evidence on AI-driven speedups comes from lagging indicators. Its version of the key sentence read: “As a result, we may not identify a significant acceleration in our rate of development until after it has occurred.”
The amendment swaps “may not identify” for “have in the past not identified, and may in the future not identify.” A forecast became an admission.
The “imminently” sentence and the human-control pledge are new as well, while the phrase “recursive self-improvement” was already in the October filing.
Closing the Loop
Anthropic calls AI-assisted research “closing the loop,” and it pitches it as an advantage. Per the filing, Anthropic’s models now run experiments, generate and process training data, automate evaluations and safety audits, and write and review code for the models that follow them.
The usage numbers are big. The value of Anthropic’s internal use of its own models, at list prices, grew more than 250 times from January to September 2026. In April, that monthly value passed what the company spends on its people each month, stock pay excluded.
The same risk factor runs the other way too. If Anthropic leans on AI too little, it says rivals could leave it “falling progressively further behind.”
Safety Plan and Human Control
Anthropic’s answer is to keep people in charge. It doesn’t expect to be able to safely let its AI “fully autonomously design and build successive generations of frontier models without meaningful human supervision or understanding, if and when that becomes technically possible.” It’s considering policy updates “to keep humans in control of frontier model development at Anthropic.”
The bar is specific. Version 3.4 of the Responsible Scaling Policy, in effect since July 8, counts one path to the threshold as double the rate of AI progress, which it defines as “as much progress in one year as one would see in two years at baseline.”
Anthropic has hit the brakes before. The filing says its precautions “have included delaying the release of models and limiting their initial availability, such as with Mythos through Project Glasswing.”
Why It’s in an IPO Filing
This section is mandatory. Item 105 of Regulation S-K requires “a discussion of the material factors that make an investment in the registrant or offering speculative or risky.” Under Section 11 of the Securities Act, buyers can sue over a material misstatement or omission in a registration statement.
So when Anthropic’s internal estimates point at the threshold, the prospectus has to say it. Buyers of the stock won’t control the board, though. The Long-Term Benefit Trust, a panel of trustees with no financial stake, elects directors who make up a majority of the board. A single Class F share held by the seven co-founders’ LLC carries 50.1% of the votes on most other stockholder decisions.
The prospectus opens with a letter signed by all seven, illustrated with art like this.
The word to watch is “imminently.” Anthropic put it in a document it can be sued over, right next to an admission that it missed a speedup once already. That’s a stronger signal than any keynote.
