Anthropic outlines three internal metrics that AI companies can use to monitor the pace and characteristics of model development. The company says it measures how much AI-led research and development is driving progress, how much oversight is applied to AI agents, and how compute resources are allocated within the organization.
The outlets frame the proposal as an effort to encourage broader transparency around how advanced AI systems are built. CNBC focuses on the specific categories Anthropic uses—AI-driven R&D, agent oversight, and compute allocation—while the Times of India describes the same three metrics as a mechanism that companies can adopt to document what their models can do and how safety and human control are handled.
Both sources describe the proposal as a template for other AI labs to follow. The Times of India additionally highlights Anthropic’s desire for shared transparency data beyond Anthropic, aiming to narrow information gaps between developers and the public.
