The White House has set out how it wants to govern the most capable AI systems, and the answer is cooperation rather than control. An executive order signed on June 2, titled Promoting Advanced Artificial Intelligence Innovation and Security, frames the federal role as a partner to industry rather than a gatekeeper over it. The order works two tracks at once: hardening the country's cyber defenses against AI-enabled attacks, and building a voluntary process for vetting powerful new models before they ship.
The clearest signal sits near the end of the text. The order states that nothing in it should be read to authorize "any mandatory governmental licensing, pre-clearance, or permitting requirement" for the development, publication, or release of an AI model. For the labs that spent the past two years worried about a licensing regime, that line is the headline.
A clearinghouse for vulnerabilities
On the security side, the order sets a tight clock. Within 30 days, the Treasury Secretary, working with the National Cyber Director, the NSA, and CISA, is directed to stand up an AI cybersecurity clearinghouse. Think of it as a shared workbench where government and industry coordinate scanning for software flaws, validate the ones they find, and push out patches without tripping over each other. Participation is voluntary, and it pulls in operators of critical infrastructure alongside the AI companies themselves.
The Department of Homeland Security is told to issue binding operational directives that open AI-enabled defensive tools to federal, state, and local agencies. The Office of Management and Budget and the Office of Personnel Management get the job of finding the money and the people, including new hiring pathways for cyber talent.
Early looks at "covered frontier models"
The second track is where the policy gets interesting. The order calls for a voluntary framework under which a developer could ask the government whether a model still in progress counts as a "covered frontier model." If it does, the company could grant federal evaluators access for up to 30 days before a wider release, under confidentiality and security protections, and help choose the trusted partners who get early access. A classified benchmarking process would sit underneath all of it to test models for advanced cyber capabilities.
It is a neat inversion of the licensing model. Rather than asking permission to release, a lab volunteers a preview in exchange for a security review and a closer relationship with the government. The order also tells the Attorney General to prioritize prosecutions of people who use AI to break into systems or steal data.
How it sits next to Congress
The timing is worth noting. Two days after the order, a bipartisan pair in the House released the discussion draft of the Great American AI Act, which would build a statutory federal framework and freeze conflicting state laws for three years. The executive order and the draft bill point in a similar direction, toward a single federal touchpoint and away from a patchwork of state rules, but they get there by different routes. One is policy the administration can act on now. The other needs votes.
Legal teams reading the order, including analysts at Hogan Lovells and Covington, read it as a deliberate bet that voluntary cooperation moves faster than rulemaking. Whether the labs take the government up on a 30-day preview, and whether that preview stays voluntary, is the question the next year will answer.
Commentarii · 0