Trump signs scaled-back executive order on AI oversight

Overview of the Executive Order

President Donald Trump signed an executive order on Tuesday establishing a voluntary framework for the federal government to vet and review powerful new artificial intelligence models before they are released to the public,. The order seeks to address national security considerations and cybersecurity threats posed by advanced AI capabilities,.

Under the terms of the order, certain AI companies are asked to voluntarily submit their new models to the government for testing or evaluation up to 30 days before public release,. The document explicitly states that the program is not mandatory and will not be construed as a federal licensing or pre-clearance requirement,.

Reductions from Previous Drafts

The final version of the order is narrower than previous iterations,. A prior draft had called for a voluntary review period of up to 90 days in advance of release, but this was reduced to 30 days in the signed version,. According to industry insiders, some members of the AI sector had pushed for a window closer to two weeks.

Trump had been scheduled to sign the more demanding 90-day version of the order in late May, but the signing was delayed. The executive order follows a postponement last month of a separate measure intended to address AI-driven cyber threats.

National Security and Industry Context

The White House described the move as a response to the fact that advanced AI capabilities introduce national security considerations requiring coordinated action across executive departments and agencies. The administration indicated it will work closely with industry to ensure the best outcomes.

This policy shift follows internal debates within the Trump administration between factions seeking to ease regulatory burdens to outcompete China and those concerned that powerful new models could undermine sensitive computer systems.

The order comes amid ongoing discussions between the federal government and leading AI firms, including Google, OpenAI, and Anthropic,. In April, Anthropic stated it was limiting the release of its Claude Mythos model due to concerns that the AI was too adept at identifying software and cybersecurity vulnerabilities. Anthropic reported last week that it would soon begin releasing Mythos-level models to customers.