TL;DR
  • President Trump signed the "Promoting Advanced Artificial Intelligence Innovation and Security" executive order on June 2, 2026
  • Creates a voluntary framework for AI companies to share frontier models with the government up to 30 days before release for cybersecurity assessment
  • Explicitly prohibits mandatory licensing or preclearance — participation is opt-in, not compelled

President Donald Trump signed the "Promoting Advanced Artificial Intelligence Innovation and Security" executive order on June 2, 2026, establishing the first formal US government framework for reviewing the most powerful AI systems before they reach the public. The order focuses on cybersecurity risks — not broad AI safety or alignment — and is designed to keep the US competitive with China while protecting critical infrastructure.

The signing came after a dramatic last-minute postponement on May 21, when Trump declined to sign a previously planned version, reportedly concerned it could "get in the way" of competing with China. The final version narrows the government review window from the originally proposed 14–90 days to a maximum of 30 days, and makes voluntary participation explicit throughout.

30 days Maximum pre-release government review period
Voluntary Company participation — not mandatory
60 days Deadline to design the voluntary framework

What the Order Does

The executive order has three core pillars:

1. AI Cybersecurity Clearinghouse — Within 30 days, the Treasury Department, NSA, and CISA must establish a clearinghouse in voluntary collaboration with the AI industry to coordinate vulnerability scanning, discovery, and patching at scale across the broader software ecosystem.

2. Voluntary Frontier Model Framework — Within 60 days, NSA, CISA, and NIST must design a framework allowing AI developers to voluntarily share "covered frontier models" with the government for up to 30 days before public release. Participation comes with confidentiality, IP protection, and cybersecurity safeguards. The government will use the access to assess advanced cyber capabilities and potentially provide early access to "trusted partners" to strengthen critical infrastructure.

3. Federal Cyber Defense Readiness — The order directs agencies to prepare federal systems for AI-enabled cyber threats.

For AI Companies: Participation is entirely voluntary — there's no legal penalty for declining. Companies that do participate gain confidentiality protections and the ability to designate "trusted partners" for controlled early access. Anthropic, OpenAI, and Google had already informally agreed to share models with CAISI under prior arrangements.

The Mythos Effect

The immediate catalyst for the order was Anthropic's latest model, Mythos — described as having exceptional capability to identify software vulnerabilities. Treasury Secretary Scott Bessent and the outgoing Fed Chair convened an emergency meeting with Wall Street CEOs after Mythos's capabilities became apparent. Anthropic has restricted Mythos to a small group of trusted partners, including large tech companies and banks, and on June 2 expanded that group by 150 additional organizations.

The executive order effectively formalizes a process the government and major frontier labs had been informally coordinating since Mythos's emergence — giving it legal structure, confidentiality protections, and a defined timeline.

Key Explicit Safeguard: The order states: "Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models." This language was essential to securing industry cooperation.

Timeline

Milestone Deadline Lead Agency
AI Cybersecurity Clearinghouse established 30 days post-signing Treasury / NSA / CISA
Covered frontier model criteria developed 60 days post-signing NSA
Voluntary sharing framework designed 60 days post-signing NSA / CISA / NIST
Federal cyber defense readiness plan 60 days post-signing Multi-agency

Industry Reaction

Response from the AI industry has been broadly positive. Groups like Americans for Responsible Innovation and the Alliance for Secure AI praised the order, with both calling on Congress to codify mandatory protections. The "voluntary" framing avoided triggering opposition from anti-regulation advocates — the White House was explicit that this is not oversight of all new models, framing broader mandatory review as government overreach that would have a "chilling effect on free speech and innovation."

Key Takeaways

  • Trump signed an AI safety EO on June 2 creating voluntary pre-release review for frontier models — the first formal US AI oversight framework
  • Companies may optionally share models with the government up to 30 days before release; no legal penalty for declining
  • Mandatory licensing or preclearance is explicitly prohibited — innovation-first framing preserved
  • Anthropic's Mythos model and its advanced cybersecurity capabilities were the direct policy catalyst
  • Industry and advocacy groups broadly supportive; Congress urged to codify mandatory protections next