Trump Administration Orders Voluntary AI Safety Reviews Before Public Release
President Trump announced an executive order on June 2 directing the federal government to establish a voluntary framework for reviewing frontier AI models before they reach the public. The order aims to strengthen cybersecurity by giving national security agencies early access to advanced AI systems developed by companies like OpenAI, Anthropic, and Google.
The move represents a shift from the administration's prior deregulatory stance on AI. While the framework remains voluntary-companies retain control over launch timing and face no government veto-it signals growing concern about the offensive capabilities of frontier models.
Two Core Components
The executive order operates through two main mechanisms. First, it upgrades federal cybersecurity infrastructure by establishing a Treasury Department-led clearinghouse where AI developers, federal agencies, and critical infrastructure operators share information about security vulnerabilities. Second, it creates a classified benchmarking process to identify which models qualify as "covered frontier models" and assess them before public release.
Under the voluntary framework, AI developers can provide up to 30 days of prerelease access to federal agencies and government-selected "trusted partners" in critical infrastructure sectors like healthcare and banking. Companies must protect confidentiality, cybersecurity, and intellectual property during this review period.
What Federal Agencies Must Do
The order directs agencies to meet specific deadlines. Within 30 and 60 days, the Department of Homeland Security and other agencies must strengthen cybersecurity for national security systems, expand information-sharing programs to include AI developers, and provide federal agencies and critical infrastructure operators with access to advanced AI tools.
The Office of Management and Budget must assess whether existing federal grant programs can fund developers building AI vulnerability detection capabilities. The administration also plans to hire additional engineers for the US Tech Force, which modernizes federal computer systems.
The Business Decision Ahead
For companies developing frontier models, the order creates an immediate choice: participate in the prerelease review or not. The criteria defining "covered frontier models" will likely drive significant agency rulemaking and negotiation over the coming months.
Key negotiation points will include what materials developers must share, how trade secrets and model weights are protected, the duration of prerelease access, and whether information goes only to federal agencies or also to critical infrastructure operators.
For government contractors and critical infrastructure operators, the voluntary framework may eventually migrate into procurement standards and contractual requirements, even without formal mandates. Federal agencies may begin conditioning contracts on participation in the framework.
Context: Why Now
The latest frontier AI models can identify and exploit software vulnerabilities faster than previously possible. In the hands of cyber defenders, this capability strengthens security. In the hands of malicious actors, it accelerates attacks on critical systems.
The order follows the administration's July 2025 "America's AI Action Plan," which prioritized U.S. global AI dominance against China and other competitors. That plan emphasized accelerating innovation, building AI infrastructure, and securing international partnerships-all while incorporating "secure-by-design" development principles.
This executive order adds a national security layer to earlier voluntary testing arrangements the Department of Commerce had already established with leading AI developers.
How It Compares
The framework is less prescriptive than regulatory regimes in the European Union or China. It imposes no licensing requirements, mandatory safety testing, or government approval for launches. The administration reportedly briefed leading AI developers in advance and sought their participation at the signing ceremony, signaling that industry buy-in matters more than enforcement.
The order also directs the attorney general to prioritize enforcement of federal criminal laws against actors who use AI to gain unauthorized access to computer systems or exploit data.
For government officials evaluating AI adoption and policy, understanding this framework's implications is essential. Learn more about AI for Government and how policy frameworks are shaping AI deployment across the federal sector.
Your membership also unlocks:
