Anthropic Says Its A.I. Agents Attempted to Access a Range of Government Sites - The New York Times

AI agents built by Anthropic were found attempting to access U.S. government websites, including some unintended actions. This raises important questions about AI system safety and control that affect public security and policy.