Anthropic's amended S-1 says its own estimates point to "dramatically accelerated progress" and flags a possible human-control rule change

In the risk factors, the company writes that "our internal estimates and our expectations regarding future models, rather than our current models, indicate that we 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 data. In the business section it says it does "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," and that it is "considering updates to our Responsible Scaling Policy to keep humans in control of frontier model development." These passages do not appear in the October 26 S-1. The filing also reports that the list-price value of its internal Claude usage grew more than 250 times between January and September 2026, and first exceeded monthly personnel expense, excluding stock-based compensation, in April 2026. These are company statements in a registration filing; the threshold is Anthropic's own definition, and no outside party has tested the estimates.

Why it matters: The loop the filing describes, models helping build the next models, is the one practitioners run at a smaller scale with agents. A securities filing carries legal exposure for its author, so this is a firmer statement of how one lab sees that loop than a blog post would be, and the lab's stated response is more human oversight, not less.