US Orders Voluntary Pre-Release Security Testing for Advanced AI Models

June 3, 2026

Mintesinot Nigussie

Leading developers of artificial intelligence systems in the United States will be asked to voluntarily submit their most capable models for government cybersecurity testing for up to 30 days before they are released to the public, under an executive order signed by President Donald Trump.

The directive assigns federal departments including Treasury, Defense, Commerce and Homeland Security, alongside other agencies, to secure agreements with AI companies to enable the testing process.

The order builds on existing voluntary testing arrangements in which companies including OpenAI and Anthropic have already submitted models for assessment by the Department of Commerce’s AI safety and standards body.

The executive order followed a delay after an initial plan to sign it on May 21 was postponed due to concerns over certain provisions and their potential impact on US competitiveness in artificial intelligence development relative to China.

Company responses have been broadly supportive. OpenAI chief executive Sam Altman said the approach “gets the balance right”, while Google executive Kent Walker described the order as “an important step forward”.