White House · 2026-06-02 · major
Trump Signs 'Promoting Advanced Artificial Intelligence Innovation and Security' Executive Order — Voluntary 30-Day Pre-Release Model Sharing With Federal Trusted Partners, CISA Cyber Push, and DOJ Focus on AI-Enabled Crime
Twelve days after the postponement, President Trump signed the AI security executive order in revised form. AI labs are asked to voluntarily share covered frontier models up to 30 days before public release, with the government picking the trusted reviewers.
The frontier-model pre-release review the White House pulled twelve days ago is now a signed executive order, with the mandatory provisions stripped out.
What is it?
An executive order issued June 2, 2026 directing federal cybersecurity, defense, and law-enforcement agencies to harden critical infrastructure against AI-enabled attacks, and asking AI labs to voluntarily share their most powerful 'covered frontier models' with vetted government reviewers for up to 30 days before public release. This is the same order Trump postponed on May 21 after telling reporters he 'didn't like certain aspects' — the rebuild keeps the structure but drops the mandatory pre-clearance language.
How does it work?
Within 30–60 days, the Committee on National Security Systems and the Department of War (DoD) must prioritize cyber defense of federal systems, CISA must issue directives to expedite the defense of civilian federal information systems, and Treasury / NSA / CISA must stand up a voluntary AI clearinghouse for vulnerability scanning. A 'covered frontier model' framework is then drafted, under which labs may share models for up to 30 days pre-release; the government helps select the trusted partners that receive early access. The order explicitly bars mandatory licensing, preclearance, or permitting. The Attorney General is told to prioritize prosecutions under existing law against anyone who 'utilizes AI to illegally access or damage a computer.'
Why does it matter?
This is the first signed US federal AI order tied to pre-release model evaluation, and it makes participation a voluntary partnership rather than a regulatory chokepoint. Labs get to keep their release cadence; the government gets early visibility on cyber capability like the ones Claude Mythos and GPT-5.5 Cyber demonstrated. It also opens up DOJ enforcement of AI-misuse cases under existing statutes — relevant for the Mercor breach and the recent Meta AI account-takeover incident.
Who is it for?
AI labs, federal policy teams, CISOs, critical-infrastructure operators