Anthropic says Claude is now taking a much bigger role in developing the company’s next generation of AI.
According to Anthropic, Claude now “leads” 26% of its AI research and development work, up from less than 1% in March. This means the model can complete most of certain tasks from a high-level prompt, while humans continue to supervise its work.
More than 90% of Anthropic’s measured AI R&D work now involves Claude at least collaborating with researchers. However, the company stressed that Claude is not operating fully autonomously in any measured area.
Anthropic says these numbers could help track how close AI systems are getting to “recursive self-improvement” — a scenario in which an AI could autonomously help build its own successor.
The company warned that AI models accelerating their own development could make increasingly capable systems harder for humans to understand or control.
Anthropic also monitors Claude’s actions using automated systems. Of more than 1 billion AI decisions analyzed in August, about 0.002% — roughly one in 47,000 — were blocked, while around 50 high-priority cases are escalated for human review each week.
Anthropic is encouraging other frontier AI labs to publish similar metrics and allow independent verification, arguing that greater transparency could help governments and the public understand how quickly AI development is progressing.