Anthropic introduces new metrics to explain AI development and oversight
Anthropic, the AI company started by Dario Amodei, is rolling out new ways to measure how advanced AI systems really are.
Its goal? To help everyone, not just tech experts, see how AIs develop, how closely they're watched, and where research resources are going.
Anthropic hopes this will make decisions about using AI more informed and transparent.
Anthropic's Claude AL4 covers 26% R&D
In August 2026, Anthropic's AI, called Claude, was at the AL4 "leads" level for 26% of Anthropic's AI R&D work.
The company also shared that it reviews every action taken by its about 30,000 AI agents and dedicates about 12% of compute used specifically for AI-driven AI R&D was allocated to safety work.
It's encouraging other labs to use similar metrics so everyone can better track how safe and independent these systems are getting.