Trump Signs Executive Order on AI Security for Federal Networks

What Happened

President Donald Trump signed an executive order on AI security, directing United States federal agencies to work with artificial intelligence companies to protect government networks from AI-related threats. The order stops short of mandating security testing for AI systems before deployment.

Key Provisions

The executive order instructs federal agencies to coordinate with AI developers on securing critical infrastructure and government networks. The administration did not include requirements for mandatory pre-deployment safety or security evaluations of AI models, a measure that had been under discussion during the drafting process. The order represents one of the first major federal policy actions under the Trump administration specifically addressing AI and national security infrastructure.

Background

The executive order follows a period of significant policy activity around artificial intelligence at the federal level. The Biden administration had issued its own AI executive order in October 2023, which included provisions requiring developers of large AI models to share safety testing results with the government before public release. The Trump administration rescinded that order in January 2025 upon taking office, directing agencies to review and revise federal AI policy with a stated emphasis on reducing regulatory burden and promoting American AI competitiveness.

The question of mandatory AI security testing has been a point of ongoing debate among technology companies, national security officials, and policymakers. Advocates for mandatory testing have argued that AI systems integrated into sensitive networks require independent evaluation before deployment. Opponents, including a number of major AI developers, have raised concerns that prescriptive testing requirements could slow development and place compliance burdens on companies.

Scope of the Order

The directive focuses on the use of AI within federal agency operations and on the protection of government networks against adversarial AI-enabled attacks. It tasks agencies with identifying vulnerabilities in their existing infrastructure that could be exploited through AI-powered tools. The order also calls for cooperation between agencies and private sector AI firms, though the specific mechanisms and timelines for that cooperation were not detailed in available reports.

The decision to exclude mandatory testing requirements distinguishes this order from frameworks adopted in some allied nations. The European Union's AI Act, which entered into force in 2024, includes risk-based requirements for certain categories of AI systems, including those used in critical infrastructure. The United Kingdom has pursued a voluntary code of practice for frontier AI developers on safety and security evaluations.

Industry and Government Context

The United States federal government has been expanding its use of AI tools across agencies including the Department of Defense, the intelligence community, and civilian departments. That expansion has prompted recurring questions about how agencies assess the security and reliability of AI systems they procure or deploy. The order arrives as several major AI developers, including companies supplying models to government contractors, are operating under existing voluntary commitments to the White House on safety practices, commitments that were established under the prior administration.

No specific AI companies were named in connection with the order in available wire reports. The order did not address export controls on AI technology, semiconductor supply chains, or the conduct of foreign AI developers operating in the United States, all of which have been subjects of separate federal actions.

What Happens Next

Federal agencies are expected to begin implementation planning in the weeks following the signing, with agency-level compliance timelines and specific coordination frameworks with industry to be established through subsequent guidance from relevant departments.