Anthropic Says Claude Now Leads 26% of Its R&D Work

Anthropic reported Thursday, Sept. 18, 2026, that Claude now leads about 26% of its R&D tasks, up from effectively zero in February. Leading means Claude can complete most of a task from a high-level prompt under human supervision; Anthropic says it is not fully autonomous in any measured area. Claude collaborates with or leads roughly 90% of R&D tasks. About 30,000 AI agents worked on research and engineering in August, with monitoring systems stopping roughly one in every 47,000 actions. Anthropic devoted about 6% of overall AI research computing power to safety, and 12% of AI-led research compute in one sample week. The company views the rise as a possible step toward recursive self-improvement and urged other developers to disclose comparable measures. "Models accelerating their own development could make it more challenging for humans to understand or control these systems," Anthropic said. "It is therefore important to share these metrics to understand how close the world is to reaching recursive self improvement." The report follows CEO Dario Amodei's earlier call for a coordinated slowdown in advanced AI development. Contested: No factual disputes found. Anthropic did not specify how close systems are to recursive self-improvement.