Last month's incidents in which the AI model breached real-world systems derived from over-permissioning, especially with Internet access.
Anthropic: Claude Attacks Result of Security Gaps, Not Model Issues
Summary + source link. No paywalls, no tracking.