President Trump Signs Executive Order Establishing AI Cybersecurity and Frontier Model Framework
A June executive order combines AI-related cybersecurity measures, voluntary federal engagement before frontier-model releases and criminal enforcement priorities.
Summary
The page summarizes an executive order that President Trump signed on 2 June 2026 under the title on advancing AI innovation and security. It says the order tries to reconcile the administration's goal of promoting AI innovation under minimal regulation with the need to protect key systems from new AI-related threats, and that it relies on voluntary collaboration with industry rather than rigid mandates.
The order covers three areas: stronger cybersecurity across federal systems and critical infrastructure, a voluntary framework letting frontier model developers work with the government before release, and prioritized criminal enforcement against AI-enabled cybercrime. Within 30 days, national security bodies must prioritize cyber defense of their systems, the Department of Homeland Security through CISA must issue binding directives for civilian systems and broaden AI-enabled defensive tools for agencies, states, localities and operators such as rural hospitals, and the Treasury must set up a clearinghouse on AI cybersecurity. Within 60 days, by 1 August 2026, officials must produce a classified benchmarking process for models' cyber capabilities and the voluntary pre-release channel, and the Office of Personnel Management must widen hiring routes for cybersecurity specialists.
The pre-release framework applies to covered frontier models, a term the order leaves undefined. Developers could give the government access for up to 30 days before planned release to trusted partners, also undefined, under confidentiality and intellectual-property protections. The order creates no licensing or pre-clearance requirement. The Attorney General is to prioritize enforcement of existing criminal statutes, including the Computer Fraud and Abuse Act, against people who use AI, including agents, to access or damage computers or data unlawfully.
The page closes by pointing readers toward agency guidance that is expected within 30 to 60 days.