Trump’s New AI Order: Voluntary Security Reviews and a Cyber Clearinghouse for Frontier Models

by Zaira Varghese 09. June 2026 AI 0

On June 2, President Trump signed an executive order titled “Promoting Advanced Artificial Intelligence Innovation and Security,” staking out the federal government’s most concrete position yet on how the most powerful AI systems should be vetted before they reach the public. The order’s defining feature is also its most debated: it is almost entirely voluntary.

What the order actually does

At its core, the order directs federal agencies to build a framework for the secure deployment of frontier AI models. The most eye-catching provision asks developers to voluntarily give the government early access to new models — for up to 30 days before wider release to trusted partners — so their capabilities and risks can be assessed. In effect, it creates a pre-release window for government review without forcing anyone to participate.

The order is careful to draw a line it will not cross. It explicitly states that nothing in it should be construed to authorize a mandatory licensing, preclearance, or permitting requirement for developing, publishing, or releasing AI models. That distinction — encouragement rather than gatekeeping — is the order’s central design choice.

An AI cybersecurity clearinghouse

Within 30 days, the order tasks the Treasury Secretary — working with the National Cyber Director, the NSA, and CISA — with forming an AI cybersecurity clearinghouse. In voluntary collaboration with industry and critical-infrastructure operators, the clearinghouse would coordinate vulnerability scanning, validate discovered flaws, and prioritize the distribution of patches. The idea is to pool detection and remediation so that AI-discovered (or AI-exploitable) vulnerabilities don’t fall through the cracks between agencies and companies.

Defining a ‘covered frontier model’

Within 60 days, the same security agencies, in consultation with NIST and the White House science office, must develop a classified benchmarking process to assess the advanced cyber capabilities of AI models — and determine the threshold at which a model becomes a “covered frontier model.” That benchmarking matters: it is the mechanism that decides which systems are powerful enough to warrant heightened scrutiny, and keeping it classified means the exact bar will not be public.

Why it’s drawing scrutiny

Reaction has split along predictable lines. Supporters argue a voluntary, capability-based framework avoids freezing a fast-moving field under rigid licensing while still giving national-security agencies a structured look at the frontier. Critics counter that “voluntary” oversight has limited teeth: if participation is optional, the developers posing the greatest risk are precisely the ones who can decline. Several legal analyses published in the days after the signing flagged that the order leans heavily on cooperation it cannot compel.

There’s also the question of capacity. Standing up a cross-agency clearinghouse and a classified benchmarking regime in 30-to-60-day windows is aggressive, and the order pulls in Treasury, the NSA, CISA, Commerce, and NIST — agencies that will have to coordinate quickly to hit those deadlines.

The bigger picture

The order lands amid a frenzied stretch for AI: record capital raises, a data-center buildout straining the power grid, and labs publicly debating how fast is too fast. Against that backdrop, the administration is signaling that it wants visibility into frontier systems without throttling them — betting that cooperation and benchmarking can substitute for regulation.

Whether that bet holds depends on the next two months. If major labs opt in and the clearinghouse delivers real coordination, the voluntary model could become a template. If the biggest players sit it out, the order risks being remembered as a framework that asked nicely and hoped for the best. Either way, the federal government has now put a stake in the ground on frontier AI — and the deadlines it set will tell us a lot about how serious the follow-through is.