Saved report copy
Recorded
Anthropic warns it is close to recursive self-improvement
Anthropic disclosed in an updated S-1A filing that its use of AI to accelerate research and development may soon meet its Responsible Scaling Policy's definition of dramatically accelerated progress attributable to AI. The company said no current model has crossed that threshold, but its internal estimates suggest it may imminently do so, and it plans to keep humans in frontier model development. Recursive self-improvement could let AI build more capable AI faster than safety and alignment measures can keep up, a risk Anthropic says it is monitoring.