Trump Signs AI Security Directive, Replacing Biden Policy and Expanding to Multiple Vendors
Updated
Updated · The White House · Jun 5
Trump Signs AI Security Directive, Replacing Biden Policy and Expanding to Multiple Vendors
3 articles · Updated · The White House · Jun 5
Summary
A new National Security Presidential Memorandum orders U.S. defense and intelligence agencies to speed AI adoption, onboard advanced models from multiple vendors, and build high-security computing capacity for military and intelligence use.
The directive says fielded systems must be robust, steerable and controllable, keeps commanders and agency heads accountable, and bars any commercial or other entity from disabling or altering mission-critical AI without prior approval.
It also creates an AI National Security Strategic Reserve of outside experts and tells the Secretary of War to update autonomy-in-weapons guidance, with annual reviews across the national security enterprise.
The memo rescinds Biden-era NSM-25, which the administration says slowed adoption and encouraged single-vendor dependence; the shift follows May agreements with eight AI companies to deploy tools on classified Pentagon networks.
As the US accelerates military AI, how will it prevent a new global arms race?
How can AI accountability be enforced within top-secret programs shielded from public oversight?
Trump’s 2026 AI Executive Order: Voluntary Security Reviews, Deregulation, and the New U.S. Cybersecurity Framework
Overview
On June 2, 2026, President Donald Trump signed an executive order called 'Promoting Advanced Artificial Intelligence Innovation and Security,' marking a new era for U.S. AI policy. The order aims to speed up the adoption and secure use of advanced AI, especially in national security and critical infrastructure. The administration chose a deregulatory approach to keep the U.S. competitive and avoid slowing innovation with strict government rules. A key feature is a voluntary 30-day pre-release review for 'frontier' AI models, letting developers submit their models for security checks before public release, balancing innovation with proactive security.