Anthropic says Claude now leads 26% of its AI research work

Anthropic reported Thursday that its AI model Claude now leads 26% of the company's AI research and development work, a sharp increase from less than 1% in February, signaling a major shift toward AI developing AI. More than 90% of Anthropic's AI R&D is now performed at a level where AI either collaborates with humans or leads the work, according to a company blog post. However, Claude has not achieved fully autonomous operation in any measured area, and the company created an index to track these automation levels. Anthropic also disclosed that about 30,000 AI agents conduct research and engineering work on its internal platform, with all actions monitored online and offline. The company allocated roughly 6% of computing power to safety work, and it plans to publish these measurements regularly with independent third-party verification.