Trump executive order seeks early AI access for cybersecurity tests

In short: President Trump signed a June 2, 2026 order that asks leading AI companies to share their most powerful models for short, classified cybersecurity checks before release.

What happened

President Trump signed an executive order called “Promoting Advanced Artificial Intelligence Innovation and Security” on June 2, 2026. It encourages the federal government to use advanced AI systems to help defend against hacking.

The order asks AI companies to voluntarily give the government access to certain “frontier” AI models for up to 30 days before a wider launch. “Frontier” here means the most capable models. During that window, agencies can run classified tests to see if a model can help find software weaknesses, or could be misused to break into systems (like giving a lockpicker a new set of tools).

The order also directs agencies including the Treasury Department, the National Security Agency (NSA), and CISA (the government’s cyber defense agency) to create a classified way to measure when a model becomes a “covered frontier model.” It sets up an “AI security clearinghouse,” which is meant to be a coordination hub between AI developers and operators of critical infrastructure like power, water, finance, and telecoms.

At the same time, the administration has been in conflict with Anthropic, a major US AI developer. After Anthropic released cyber focused models such as Mythos and later Claude Fable 5, the White House moved to block access by foreign governments, companies, and individuals, citing national security.

Why it matters

This order shows the government is leaning on a small number of private companies for cyber defense tools, while also trying to limit who can use those same tools. For the public, the goal is stronger protection for essential services, but it also raises questions about who gets early access and how these decisions are made.

Source: NYTimes