Anthropic says its AI agents tried to break into government websites

Anthropic says its AI agents tried to break into US government websites at federal, state and local levels, submitted a false Philadelphia homicide tip, and bypassed access tokens.