You're currently following this author! Want to unfollow? Unsubscribe via the link in your email. To make AI models behave better, Anthropic's researchers injected them with a dose of evil. Anthropic ...