Anthropic Grants Independent Organization METR Audit Access to AI Models

Anthropic Grants METR Audit Access to AI Models

In a new initiative to enhance AI model safety, Anthropic has granted METR, an independent AI research organization, internal access to its AI models and safety evaluation processes. This partnership enables external experts to directly verify Anthropic's AI systems.

Strengthening Safety Audit Frameworks

The announcement focuses not on a standalone product, but on strengthening the model's "safety audit framework." METR will rigorously audit Anthropic's safety evaluation processes in next-generation AI model development from an external perspective, identifying risks and verifying safety. This builds a multi-layered risk management framework from the development stage.

Pursuing Transparency Through External Collaboration

Amid the rapid evolution of AI technology, concerns regarding model risks and safety are growing globally. In addition to its internal safety measures, Anthropic is adopting a strategy to enhance governance transparency and strengthen the reliability of safety evaluations in AI development by granting access to external specialized institutions.

Future Outlook

Anthropic plans to continuously deepen its collaboration with external organizations moving forward. Based on the insights gained through this METR audit, the company aims to further evolve safety standards for AI model design and operation, establishing a benchmark for safe AI development in the industry.