TechAnthropic Reports Claude AI Model Now Leads 26% of Company's R&D Work
Anthropic disclosed that Claude models now lead approximately 26% of its internal AI research and development work as of August 2026, up sharply from under 1% in March. The company runs ~30,000 agents concurrently with ~6% of compute dedicated to safety work.
TLDR
This represents one of the clearest public signals yet of frontier AI accelerating its own development. It fuels excitement about capability gains alongside urgent safety and alignment conversations regarding agent oversight and misuse risks. Anthropic positions itself as a transparency leader by proposing standardized metrics for tracking AI-driven R&D pace and urging other labs to adopt similar reporting.