EO-14409
Promoting Advanced Artificial Intelligence Innovation and Security
- Signed
- Jun 2, 2026
- Published
- Jun 5, 2026
Federal Register: 2026-11415
Source: Federal Register.
AI Innovation and Cybersecurity Hardening Order
What it does
This order directs multiple federal agencies to strengthen cybersecurity defenses across government, critical infrastructure, and private sector systems using AI-enabled tools, all within 30–60 days. It establishes a voluntary framework for AI developers to share advanced "frontier" AI models with the government before public release, and creates a clearinghouse to coordinate the discovery and patching of software vulnerabilities. It also directs the Attorney General to prioritize prosecution of crimes that use AI to illegally access computer systems.
Who benefits
Federal civilian agencies receiving upgraded cybersecurity tools and guidance; rural hospitals, community banks, and local utilities that would gain access to government cybersecurity services; AI developers who gain early government feedback on their models without mandatory licensing requirements; state and local governments receiving expanded cybersecurity support; cybersecurity professionals and job-seekers through expanded federal hiring pathways; critical infrastructure operators who benefit from coordinated vulnerability patching; victims of AI-assisted computer crimes who may see stronger federal prosecution.
Who is affected
AI developers and researchers who may face informal pressure to participate in the "voluntary" pre-release review framework despite its stated non-mandatory nature; smaller AI companies with fewer resources to engage in government review processes, potentially disadvantaging them relative to large incumbents; civil liberties advocates concerned about government access to AI model internals; foreign AI competitors who may face a more coordinated U.S. government-industry AI security posture; federal employees whose agencies must rapidly implement new cybersecurity directives within tight 30-day windows; taxpayers who fund the expanded programs, clearinghouses, and hiring initiatives.
Supporters argue
Supporters argue that coordinating AI cybersecurity across government and critical infrastructure fills a genuine gap — adversaries are already using AI to probe and attack U.S. systems, and a fragmented federal response leaves hospitals, banks, and utilities exposed. They contend the voluntary framework for frontier model review strikes the right balance: the government gains early awareness of potentially dangerous AI capabilities without imposing the mandatory licensing or preclearance that would slow innovation and chill research, as Section 3(c) explicitly prohibits.
Opponents argue
Opponents argue that a "voluntary" pre-release review framework backed by government resources and relationships creates de facto pressure on developers to comply, blurring the line between voluntary and mandatory without the procedural protections that formal regulation would require. They contend that classified benchmarking criteria for designating a model a "covered frontier model" — criteria developers cannot publicly scrutinize or challenge — raises due process concerns about vague and opaque standards, and that expanding NSA and CISA access to AI model internals risks normalizing government surveillance of private technology development.
Constitutional basis
Executive orders rest on constitutional authority or statutory delegation. This summary describes the legal grounding cited or implied by the order.
The order rests on the President's Article II authority as Commander in Chief and chief executive over the federal bureaucracy, directing agencies within the executive branch. It also draws on statutory authority delegated by the Federal Information Security Modernization Act (FISMA, 44 U.S.C. § 3552), which authorizes Binding Operational Directives for federal cybersecurity, and the Computer Fraud and Abuse Act (18 U.S.C. § 1030) and related statutes cited in Section 4 for criminal enforcement prioritization. Because the order explicitly prohibits mandatory licensing or preclearance of AI models (Sec. 3(c)), it avoids the most direct First Amendment and major-questions-doctrine conflicts, though the classified benchmarking process and voluntary-but-government-facilitated review framework could face scrutiny under post-Loper Bright (2024) heightened review of agency statutory interpretations.