Anthropic CEO Dario Amodei has called for greater transparency around AI development, including public reporting on how models are built and used.
His proposal reflects a wider debate over the pace of AI development, as capabilities advance while the public and policymakers have limited visibility into how models are trained and improved.
Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram
The call has drawn support from OpenAI CEO Sam Altman, Tesla and SpaceX CEO Elon Musk, and Google DeepMind Chair Demis Hassabis. Amodei has said any slowdown should preserve commercial competitiveness and the United States’ lead in AI.
Anthropic Tracks AI Development With Three Metrics
Anthropic said its first metric found that Claude models were not fully autonomous in any subset of the research and development work it measured.
A second metric found roughly 30,000 AI agents performing research and engineering work across the company’s most-used internal platform, with a system in place to monitor and intervene in their actions.
For the third metric, Anthropic examined compute use from July 13 to July 20. About 6% of compute used for AI research and development went to safety, while safety-related work accounted for roughly 12% of compute dedicated to AI-driven research and development.
Why The Metrics Matter
Anthropic said the measures complement capability evaluations by showing more about how AI systems are developed, rather than only what they can do.
The company said publishing the data could give outside observers a clearer basis for assessing the pace of AI development and added that it plans to continue releasing the measurements.







