Politico: 'It's a Hurricane Warning' — US Races to Vet Frontier AI for Cybersecurity Threats Before China Closes the Gap

Politico reports US officials and security experts are warning that guardrails on cyber-capable frontier AI models may arrive too late as China narrows the gap. A Trump executive order reportedly directs federal agencies — NSC, Treasury, and Defense Department — to conduct classified vetting of frontier models up to 30 days before public release, testing for offensive cybersecurity capabilities. Both the US and China are actively probing frontier AI models for offensive cyber potential; Anthropic previously accused Beijing of using Claude to automate parts of an espionage campaign targeting roughly 30 global organizations.