Anthropic has told the SEC that its internal estimates indicate it “may imminently meet” the threshold in its Responsible Scaling Policy for dramatically accelerated, AI-driven progress. TechCrunch reports the language was added to the risk factors in an amended S-1 filed on Monday.
The filing also says Anthropic has in the past not identified major capability accelerations until after they happened. TechCrunch’s Julie Bort reads that as an admission that AI-assisted research has already produced leaps the company missed at the time.
Anthropic says it has not designated any current model as past the threshold, and that it continues to assess the data. This account rests on TechCrunch’s excerpts, not the full filing. It says it does not expect to use Claude to design successive frontier models without meaningful human oversight, and that it is considering updates to the policy to keep humans in control.
Considering is not adopting. No change has been announced, and the filing gives no figure for how fast progress is moving or what imminently means in months.
A sceptic cited by TechCrunch argues that an S-1 is both disclosure and marketing. Bort’s own view is that if Anthropic believes self-improving AI is close, readers might as well take it at its word. The distinction that matters is between an AI that helps researchers and one that replaces them, and the filing does not say which the company expects to cross.
Risk factors exist to protect a company from later claims that it hid something. The open questions are which past models the filing means, and what Anthropic would do on the day it says the threshold has been crossed.
