Anthropic unveils 'auditing agents' to test for AI misalignment
3,43K