Trump AI Executive Order To Seek Early Access To Advanced Models

After Anthropic's 'Mythos' model sent shockwaves through the cybersecurity world due to its ability to find and exploit software vulnerabilities at breakneck speed, the Trump administration is reportedly on the cusp of issuing a much-discussed executive order that would encourage AI companies to provide information on their advanced models to the government before public release. 

According to Axios, the order - which could come as soon as this week - will outline plans for a voluntary framework - meaning companies can just ignore it - under which AI labs would share their models with the government at least 90 days before public release, while also giving access to certain critical infrastructure providers. 

Mythos and OpenAI's latest model, GPT-5.5-Cyber, have raised alarm bells both inside and outside government due to their ability to find and exploit software vulnerabilities with unprecedented speed. 

The EO will also cover cybersecurity, and "aims to secure the Pentagon and other national security agencies, boost cyber hiring, shore up cybersecurity systems across the country at places like hospitals and banks, and encourage threat sharing about breaches between the AI industry and government."

The component covering the advanced 'frontier' models such as Mythos would involve multiple layers of government review to see if it qualifies as a "covered frontier model," and then assess them prior to public release.

The voluntary 90-day pre-release sharing framework lets the government:

• Review models early via national security and civilian agencies.
• Assess risks.
• Advise labs or critical infrastructure providers.
• Prepare defenses if needed.

This gives the White House situational awareness on what's coming down the pike, without trying to outright regulate or slow U.S. AI companies (which would contradict the administration's "America first / global dominance" stance on AI).

In short: It signals "we want to keep an eye on the dangerous models" to the public and adversaries, builds relationships for threat intel, and keeps the U.S. competitive. Whether companies actually engage will depend on norms, pressure, and self-interest. If the final version (expected soon) adds more carrots/sticks, its teeth could sharpen.