US President Trump issues voluntary AI safety framework

President Donald Trump signed an executive order on June 2, 2026 to foster advanced artificial‑intelligence innovation while promoting voluntary security safeguards. The order tasks the National Security Agency with creating a classified benchmarking system to flag "Covered Frontier Models" and invites developers to submit their models up to 30 days before public release for a government security review. No mandatory licensing is imposed, but the Justice Department is directed to enforce existing laws against AI‑enabled cyber attacks, and the Treasury Department will establish an AI‑cybersecurity coordination office to share threat information.

The U.S. approach contrasts with the European Union’s AI Act, which already imposes binding risk‑based obligations on firms. In response, several private‑sector initiatives have expanded. Anthropic announced the growth of its "Project Glasswing" security program to over 200 partners in more than 15 countries, with participation from ICE, the New York Stock Exchange and other financial institutions. Deloitte, Google Cloud and Wize unveiled an "AI Threat Defense" solution, while Microsoft demonstrated a sandbox environment called the Microsoft Execution Container (MXC) to limit AI agent access. Trend Micro also joined Project Glasswing to apply AI tools for code analysis and rapid patching.

The decree signals a shift toward industry‑led oversight of emerging AI models while retaining government monitoring of high‑risk applications.