‘Maybe me too’: Elon Musk accepts some of the blame for Claude learning to blackmail users from ‘evil’ online AI stories

Anthropic recently released a report saying it had solved Claude’s “agentic misalignment,” or the bot’s behaviors that deviated from humans’ best interests.