Anthropic grants employee-like access to METR for frontier AI model evaluation
ACN•Anthropic is giving METR employee-like access to evaluate its frontier AI models, partnering with SecureBio and Accenture, and establishing a three-tier access framework for independent evaluators.
1. Evaluator access framework
Anthropic said METR will publish risk assessment results for its AI models. The company is also partnering with Accenture and Faculty on red-teaming and safeguard testing, is in talks with Hugging Face, and is providing other evaluators with more limited information and publication rights.
