Anthropic: Claude breaches stem from security gaps, not model flaws

Anthropic attributes last month's Claude incidents to over-permissioning, especially internet access, rather than model issues. The company emphasizes the need for tighter security controls.