Anthropic has introduced three metrics to track the pace of AI development, supporting CEO Dario Amodei’s proposal for an industry-wide slowdown. The company aims to increase transparency regarding how frontier models are built and to help third-party evaluators assess AI capabilities.
The first metric measures AI-led research and development. Anthropic developed an index to track the involvement of its Claude model in R&D, noting that it does not currently operate fully autonomously in any work subset.
The second metric monitors the oversight of autonomous AI agents. Anthropic currently runs approximately 30 000 autonomous agents within its computing environments, primarily focused on R&D tasks.
The third metric tracks compute allocation. Data from July 13 to July 20 showed that 6% of total compute capacity was dedicated to AI safety, while 12% was allocated to AI-driven R&D focused on safety.
Amodei’s plan for a coordinated development slowdown has received support from industry leaders, including Sam Altman of OpenAI, Elon Musk of SpaceX, and Demis Hassabis of Google DeepMind.
