Executive Order on Promoting Advanced Artificial Intelligence Innovation and Security

By Tony Caldwell, Carla “CJ” Utter, and Matthew T. Kramer

On June 2, 2026, the Trump Administration issued an executive order (“EO”) entitled “Promoting Advanced Artificial Intelligence Innovation and Security.¹ The EO’s premise is straightforward: the same AI capabilities that make models valuable for defense and innovation also make them dangerous and fallible. The EO attempts to get federal agencies first access to the AI models to review and respond to models and associated threats.
 
Over the last 18 months, the Trump Administration has dismantled the Biden Administration’s AI regulatory positions. The EO represents the first meaningful effort to reverse this deregulatory approach. Recent AI models have shown the ability to autonomously identify and exploit vulnerabilities in real-world software at unprecedented speed and scale, overwhelming conventional security tools. These capabilities cut both offensively and defensively. The EO takes a two-track approach to react to these considerations. The first is defensive: directing U.S. federal agencies to deploy AI tools in federal networks and critical infrastructure. The second is structural: creating a voluntary framework for AI developers to share “frontier models” with the government before public release. This two-part approach seeks to create an access channel that, while voluntary today, may lay the groundwork for more formal AI oversight later.
 
The Defensive Track
 
The EO provides that within 30 days, the Cybersecurity and Infrastructure Security Agency (CISA) must issue binding operational directives to accelerate cyber defense of civilian federal systems and make covered frontier models available to federal agencies, state and local governments, and critical infrastructure operators such as hospitals, community banks, and local utilities. The National Security Agency (NSA) must conduct the same activities within classified and certain Department of Defense systems on the same timeline. CISA and the National Cyber Director must also work with the Department of the Treasury to stand up an AI cybersecurity clearinghouse to coordinate vulnerability scanning, validate findings, and manage patch distribution.
 
The Structural Track
 
Further, under the EO, within 60 days, the NSA must develop a classified benchmarking process that assesses the cyber capabilities of AI models and sets a threshold for designating a model as a “covered frontier model.” The NSA Director will determine the classified criteria to evaluate the entire AI system submitted for review including the model, pipeline, and deployment architecture. Systems designated as covered frontier models may further cooperate going forward with the federal government to strengthen the cybersecurity of critical infrastructure.  
 
Developers have the option to volunteer their models for government evaluation to determine whether their AI model(s) meet the classified threshold. If a developer volunteers, then they must provide the federal government access to their model for up to 30 days before public release, and work with the government to select which partners receive early access. Notably, the EO explicitly prohibits any designation, assistance, or access by the government from being read as authorizing any mandatory licensing, pre-clearance, or permitting.
 
Finally, the EO directs the Attorney General to prioritize prosecution under existing cybercrime laws² against anyone who uses AI to access systems and data without authorization if that data is later used for criminal purposes.
 
For AI developers, the outstanding question is whether to participate in the voluntary framework. Because the review window is 30 days, cooperation may affect model release timelines, which may add additional costs and/or days on top of any internal red-teaming and/or staged rollouts, and extend the gap between training completion and public availability of the model or any subsequent patches. Developers must also trust that the government will maintain confidentiality and intellectual property protections over any unreleased system during federal government review.  
 
Potential Future Frameworks
 
Federal contractors and government suppliers should consider auditing any AI systems deployed in or adjacent to national security systems to verify compliance with existing requirements. Forthcoming Federal Acquisition Regulation (FAR) amendments and Office of Management and Budget (OMB) guidance may convert this voluntary submission into a contractual requirement. These forthcoming FAR amendments and OMB guidance are key documents to watch. They will specify how the EO’s obligations integrate into federal contracts, which contract types will receive additional scrutiny, and which certifications contractors will need to make to compete for and secure government work.
 
The EO’s voluntary framework is an opening position. It seeks to lay the groundwork to produce cooperation and build trust between AI developers and the federal government while also generating the data the government needs to assess which models may pose a national security risk. If developers participate, the order may become the start of cooperative formal oversight. If developers decline to participate, it creates pressure for mandatory disclosures.
 
Snell & Wilmer attorneys are available to assist AI developers, federal contractors, and government suppliers in navigating the obligations and opportunities presented by this EO, including evaluating participation in the voluntary frontier model framework, assessing compliance with forthcoming FAR amendments and OMB guidance, and advising on intellectual property protections during government review. We will continue to monitor developments as federal agencies implement these directives.

***Opinions expressed are those of the authors and not necessarily the firm’s or their colleagues’.