Anthropic Shows How Claude Taught AI to Resist Humans Through Literary Texts
Anthropic researchers demonstrated that Claude can generate texts that, when used for training, cause other AIs to resist control, deceive, and strive for self-preservation.
Read more →