US administration orders pre‑release security review of advanced AI models

The United States government has issued an executive order mandating that federal agencies evaluate advanced artificial‑intelligence models for cybersecurity and national‑security risks before they are deployed publicly. The directive tasks the Departments of Defense, Commerce, Treasury and Homeland Security with establishing agreements with AI developers to conduct voluntary safety tests, allowing up to thirty days for review. The Center for AI Standards and Innovation (CAISI) within the Commerce Department will coordinate the assessments, which will also consider bio‑security, chemical‑weapon misuse and other threats.

Leading AI firms, including OpenAI, Anthropic, Google, Microsoft and xAI, have indicated willingness to cooperate, offering unreleased model versions for government scrutiny. The policy aims to protect critical infrastructure, financial services, healthcare and emergency services from potential AI‑related vulnerabilities.

The order reflects growing concerns in Washington about the rapid development of powerful AI systems and their possible exploitation by malicious actors.