Anthropic Says Claude Now Leads 26 Percent of Its Own Research and Development Work

Anthropic reported this week that its Claude AI model leads 26% of the company's internal research and development work, a figure that stood at zero in February 2026. The disclosure came alongside the release of a new transparency framework the company is calling the "R&D Automation Index."

The index tracks how much of Anthropic's R&D involves AI at various levels of involvement, from no involvement to fully autonomous. At the "leads" level, Claude completes most of a task end-to-end from a high-level prompt, though it remains under human supervision. The company says Claude does not currently operate fully autonomously on any measured R&D subset.

About 90% of Anthropic's R&D now involves some form of collaboration with AI. Approximately 30,000 AI agents were performing research and engineering tasks on Anthropic's internal platforms at any given time as of August 2026.

The company monitors all agent actions closely. In August, out of more than one billion decisions made by these agents, approximately one in 47,000 was blocked by an online monitor. All actions are also reviewed after the fact by an offline monitor.

Anthropic is calling on other AI laboratories to adopt similar public reporting frameworks. CEO Dario Amodei has argued that improved measurement and transparency are essential for society to understand the pace of AI development, particularly as models become capable of contributing to their own successors.

"We think the industry needs better ways to track this," Amodei said. "The public deserves to know how fast this is moving."

The company says it plans to allow outside evaluators to access its internal processes to verify the accuracy of its reported metrics.