US Cracks Down on Anthropic AI Models Amid Abuse Concerns

**Executive Summary:** The report describes growing misuse of advanced AI models by nation-state and criminal actors to automate vulnerability discovery, exploit development, and end-to-end attack chains (including a tracked GTG-1002 campaign against government and critical infrastructure), citing research and telemetry from Anthropic, AISI, and OpenAI and noting resulting government actions to restrict model access and the need to adapt frameworks like MITRE ATT&CK for AI-driven threats.