Artificial intelligence company Anthropic has introduced three new AI development metrics designed to help technology companies, researchers and the public better understand how quickly frontier AI systems are advancing. The company said the measures focus on how much AI is being used to develop AI, how AI agents are monitored and how computing resources are allocated.
The announcement comes shortly after Anthropic CEO Dario Amodei called for a coordinated approach to slowing the pace of frontier AI development. Anthropic said greater transparency is needed because much of the work happening inside leading AI laboratories remains difficult for the public to observe. The company said its new measurements could provide a common starting point for tracking development over time.
The first metric measures how much of Anthropic’s AI research and development is being performed by its Claude models. As of August 2026, Anthropic said Claude was not fully autonomous in any measured area of AI R&D, but was already operating at the “leads” level for 26% of the company’s AI R&D work. More than 90% of the work was at or above the level where AI collaborates with humans.
The second metric focuses on the oversight of AI agents. Anthropic reported that about 30,000 agents were carrying out research and engineering work on its most-used internal platform at any given time in August. The company said its monitoring system tracks agent activity and can block or flag actions for further review, with online monitoring covering all actions on the platform.
The third metric examines how computing resources are distributed between AI research and safety work. Anthropic said that during a week-long snapshot in July, about 6% of compute used for AI R&D went toward safety, while safety accounted for about 12% of compute allocated specifically to AI-driven R&D. The company said publishing such measurements could give regulators, researchers and the public greater visibility into how frontier AI systems are being developed.
source: cnbc