On Monday Anthropic filed an amended version of its IPO prospectus (S-1A) with the US Securities and Exchange Commission, with updated financials. According to TechCrunch, the company added new statements to its risk factors about using AI to build AI, a process known as "recursive self-improvement" (RSI).
In the filing the company wrote that its growing use of AI systems to accelerate its research and development may cause its models' capabilities to advance faster than its ability to develop, validate and implement corresponding safety, alignment and security measures, and that in the past it has not identified a significant acceleration in its rate of development until after it occurred. It added that its internal estimates and expectations for future models indicate that it "may imminently" meet its Responsible Scaling Policy's (RSP) definition of "dramatically accelerated progress attributable to AI."
Anthropic stressed that it has not designated any of its current models as exceeding the threshold and that it continues to assess the data. The company wrote that 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 the policy to keep humans in control of frontier model development.
It is not yet known which future models the estimate refers to or exactly when the threshold might be reached. TechCrunch notes that a prospectus serves both as disclosure and as marketing.