Trump Executive Order Establishes Voluntary AI Model Review

President Trump signed an executive order on June 2, 2026, establishing a voluntary framework for federal review of cutting-edge artificial intelligence models before they reach the public market. The directive asks AI developers to submit certain advanced models to federal agencies 30 days before release, giving the government time to identify security vulnerabilities.

The executive order, titled "Promoting Advanced Artificial Intelligence Innovation and Security," represents a notable shift for an administration that initially characterized AI safety concerns as "hand-wringing" and criticized previous regulatory efforts as obstacles to competitiveness. Trump had planned to sign a similar order on May 21 but pulled back over competitiveness concerns before moving forward with this version.

Why it matters

This marks the federal government's most extensive engagement with AI markets to date, surpassing even the Biden administration's 2023 executive order that required notification and safety testing for models trained on high computing power. The voluntary nature preserves industry flexibility while addressing mounting concerns about AI-enabled cybersecurity threats to critical infrastructure. As advanced models like Anthropic's Mythos demonstrate capability to detect system vulnerabilities at scale, the framework attempts to balance innovation with security without imposing licensing requirements or giving government veto power over product launches.

Core provisions

The executive order establishes three main focus areas, according to details first reported by Crowell & Moring.

First, it directs multiple agencies to upgrade federal cybersecurity defenses. The Committee on National Security Systems has 30 days to prioritize protection of national security systems, while the Cybersecurity and Infrastructure Security Agency must release a Binding Operational Directive to enhance AI-enabled defensive tools across federal, state, and local governments. The Treasury Secretary must establish an "AI cybersecurity clearinghouse" within one month to facilitate voluntary collaboration between AI developers and critical infrastructure operators on vulnerability scanning and patch distribution.

Second, the order creates a process for designating "covered frontier models." The Treasury Secretary, Secretary of War working through the National Security Agency, and other agencies will develop a classified benchmarking process to assess AI models' advanced cyber capabilities. The NSA Director will determine which models receive this designation. Developers can voluntarily engage with government before release, provide 30-day advance access to covered models, and share access with select critical infrastructure operators.

Third, the directive instructs the Attorney General to prioritize enforcement of federal criminal laws against actors using AI for unauthorized computer access or to enhance other crimes. This builds on existing FBI and Justice Department focus on AI-enabled criminal activity, including recent discoveries of threat actors using AI to exploit previously unknown code vulnerabilities.

Regulatory landscape

The executive order explicitly states it does not create licensing, preclearance, or permitting requirements for AI model development or release. However, it arrives as AI developers face an increasingly complex compliance environment with investigations from state attorneys general, the Federal Trade Commission, Securities and Exchange Commission, congressional committees, and international bodies all examining various AI-related concerns.

The White House began meeting with industry about the order's potential details in April, shortly after Anthropic unveiled Mythos to a limited audience of technology companies and security researchers. That model's reported ability to detect cyber vulnerabilities quickly and at scale raised concerns among governments and private sector entities worldwide.

These details were first reported by Crowell & Moring.