Trump signs AI order to boost cybersecurity and allow NSA review

President Trump on Tuesday signed an executive order titled “Promoting Advanced Artificial Intelligence Innovation and Security” that directs federal agencies to expand use of AI for cybersecurity, establishes a voluntary review process for powerful models and permits the National Security Agency to conduct classified evaluations of certain systems.

The order instructs agencies to adopt AI-powered cybersecurity tools more quickly, establish an AI cybersecurity clearinghouse to share defenses, and create a process to identify so-called frontier models that may pose national security risks. The NSA will determine whether advanced systems meet the definition of covered frontier models. Developers may voluntarily provide those models for government evaluation for up to 30 days before planned release to trusted partners. The administration said the framework is not intended to create a formal approval process for model releases.

The order directs federal departments to prioritize deployment of defensive AI tools and to share information on exploits and mitigations across agencies. It set enforcement priorities targeting criminal uses of AI, including employing AI agents to unlawfully access public or private computer systems or to generate illicit content for criminal purposes.

The signing follows a pause in May, when the president delayed a previous version of the order, citing concerns that parts of the proposal could slow U.S. AI development and weaken competitiveness with China. Administration officials revised the proposal before Tuesday’s signing.

Officials raised internal alarms earlier this year after a generative model demonstrated an ability to identify software vulnerabilities. That disclosure prompted high-level meetings about cybersecurity risks and led some firms to seek stricter oversight. The developer of the model expanded a controlled-access program, adding roughly 150 organizations to test and fix potential exploits ahead of broader release.

Civil society groups and consumer advocates criticized the order’s reliance on voluntary cooperation from technology companies. J.B. Branch, AI governance and technology policy counsel at Public Citizen, urged Congress and the administration to pass federal AI legislation with enforceable safeguards, independent testing, transparency requirements and protections for workers, consumers, children and civil rights.

The order comes as states pursue their own AI rules and federal officials move to establish a national framework. Federal prosecutors recently charged two men under a law aimed at removing nonconsensual explicit images after alleged use of AI to create and distribute such images, one of the early enforcement actions linked to harmful AI-driven conduct.