Anthropic Is Banishing Its Model Evals From the Internet - Gizmodo

Anthropic is removing its model evaluation data from public access. This affects transparency about how well AI systems perform and whether they're safe, making it harder for outside researchers and the public to independently verify AI safety claims.