Trump Eyes New AI Executive Order for Early Government Review of Advanced Models

The White House could issue a new executive order on artificial intelligence as early as Thursday, signaling a major shift in how the United States oversees advanced AI systems before they are released to the public.

According to sources familiar with the discussions, the proposed order would establish a voluntary framework allowing AI companies to share powerful new AI models with the US government ahead of launch. The review period remains under negotiation, with some draft versions proposing a 90-day advance review while several AI firms are pushing for a shorter timeline of around 14 days.

Major AI developers, including OpenAI and Anthropic, have reportedly been in talks with the White House regarding the policy.

The move comes amid growing concerns that increasingly capable AI models could accelerate cyberattacks and expose critical vulnerabilities. Experts warn that frontier AI systems may enable hackers to automate sophisticated attacks at unprecedented speed.

One draft of the executive order reportedly contains two key sections: cybersecurity and “covered frontier models.” The latter would define which advanced AI systems qualify for the voluntary government review process before public deployment.

The cybersecurity section would create a voluntary “clearinghouse” involving the Treasury Department, government agencies, and AI firms to identify and address security weaknesses in unreleased AI systems. It would also expand hiring within the US Tech Force, a federal engineering initiative focused on modernizing government technology infrastructure.

The Trump administration had previously favored a lighter regulatory approach toward artificial intelligence. However, discussions intensified after Anthropic introduced its highly advanced Mythos AI model, which the company claims can rapidly exploit cybersecurity vulnerabilities. Anthropic has limited access to the system through its Project Glasswing initiative, working alongside government and private-sector partners.

Meanwhile, OpenAI has also begun granting select businesses and government agencies early access to some of its newest AI systems to strengthen cyber defenses and prepare for emerging threats.

The proposed executive order reflects growing pressure on governments worldwide to balance rapid AI innovation with national security and public safety concerns.