White House Scales Back AI Order After Industry Push

President Donald Trump signed an executive order on June 2, 2026, that asks leading AI developers to voluntarily give federal agencies early access to their most powerful models for security testing before public release.

The order creates a voluntary framework that would provide the government and select “trusted partners” with secure early access to so-called covered frontier models, and directs agencies to develop a classified benchmark to identify those systems.

Crucially, the White House text says the measure does not authorize mandatory government licensing, pre‑clearance, or permitting for the development or release of AI models — language added after industry objections.

The final directive reduces the voluntary government review window to 30 days ahead of a public release — a shorter period than earlier drafts that proposed up to 90 days — a change that reflected pushback from AI companies and advisers.

Reporting and post‑mortems say the White House postponed an earlier, tougher version of the order in May after last‑minute lobbying from industry figures and former White House AI adviser David Sacks, who argued a longer review could slow U.S. competitiveness.

The administration framed the final order around cybersecurity and protecting critical infrastructure, directing Treasury, the NSA, CISA, and other agencies to help build an AI cybersecurity clearinghouse and to prioritize defenses for national security systems.

Major AI developers already participate in voluntary testing and benchmarking arrangements with Commerce Department‑linked bodies and security partners; the EO points to expanding those cooperative channels rather than imposing new regulatory permits.

Industry groups and companies lobbied intensively to limit the potential for binding approvals or licensing, even as they signaled willingness to share benchmark results and collaborate on red‑teaming and vulnerability remediation.

Administration officials said the order balances innovation and security: it seeks early visibility into frontier capabilities while explicitly preserving industry control over release decisions. Critics caution the distinction between voluntary programs and future mandates can blur as governance frameworks evolve.

The EO also instructs the Department of Justice to prioritize enforcement against criminal uses of AI — for example AI‑assisted hacking or automated fraud — and asks agencies to expand federal hiring and funding for AI cybersecurity.

Policy analysts say the scaled‑back EO shows the leverage big AI firms and their allies have in shaping near‑term U.S. AI policy: companies secured assurances against mandatory preclearance while accepting cooperative benchmarking and cyber‑defense roles. The result is a federal posture that leans on voluntary collaboration rather than immediate coercive rules.

The administration directed agencies to stand up the voluntary programs and benchmarks on rapid timelines and to report back to the White House — a fast schedule that could still shape industry behavior even without statutory mandates. Observers say Congress and state governments remain potential next battlegrounds for harder rules.