What Just Happened

Trump signed an executive order on June 2, 2026 directing AI companies to voluntarily share powerful new models with the federal government up to 30 days before public release. The goal: let agencies assess cybersecurity risks before those models hit the market.

This is not the order that was originally drafted.

The Version That Got Killed

According to Wired, the original draft required a 90-day pre-release window. Industry executives told Wired directly that their companies weren't prepared to share models that far in advance. That provision was the most contested piece of the entire order.

Trump was scheduled to sign a broader version on May 21 — with a lineup of Silicon Valley CEOs present. Hours before the ceremony, he pulled the plug. He told reporters he didn't like "certain aspects" of it and didn't want to slow American AI down relative to China.

The Daily Signal reported that calls to Trump from David Sacks — his former AI czar — and other tech industry leaders directly preceded the cancellation.

What Sacks Did

Sacks lobbied against the executive order he had previously been in charge of shepherding. According to Wired, Sacks told Trump the order would be "too onerous" for industry. TechCrunch confirmed the delay followed "industry pushback, including from venture capitalist and former White House AI czar David Sacks."

The man who was supposed to be the administration's AI policy chief used his access to kill a national security measure. Whether you think the original order was good policy or not, that represents a conflict of interest that deserves closer scrutiny.

What the New Order Actually Does

The June 2 order, per CNBC, does the following:

• Asks AI developers voluntarily to submit frontier models for government benchmarking
• Gives the government access to those models up to 30 days before broader release
• Directs the Department of Justice to treat AI-assisted hacking and unauthorized access as high-priority enforcement
• Instructs agencies including DHS, OMB, and the National Cyber Director to prioritize AI-related cybersecurity

The order explicitly states — in its own text, quoted by both CNBC and TechCrunch — that nothing in it "shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement." That language was added specifically to reassure industry.

No licensing. No mandatory review. No permitting. Voluntary all the way down.

Why the Government Wanted This in the First Place

Both the Daily Signal and Wired report that the push for oversight was driven by a specific national security concern: frontier AI models from companies including Anthropic and OpenAI that have demonstrated the ability to identify major vulnerabilities in legacy software systems.

The White House wanted early access to these models because they can crack legacy software. That's a genuine threat vector. Foreign adversaries with access to models like that could probe American infrastructure quietly and systematically.

The administration backed away from that concern because tech executives made phone calls.

The Internal War Wired Exposed

Wired did extensive reporting revealing an administration split down the middle.

On one side: White House Chief of Staff Susie Wiles, Treasury Secretary Scott Bessent, and National Cyber Director Sean Cairncross — all pushing to resurrect meaningful oversight. Bessent reportedly met personally with Anthropic CEO Dario Amodei to find a workable path forward and is expected to lead cross-border AI negotiations with China.

On the other side: Sacks and aligned tech industry voices, arguing any oversight slows American innovation.

The faction that wanted less oversight won. The June 2 signing happened quietly — no CEOs, no ceremony, no cameras. CNBC confirmed Trump signed the order in private.

What the Coverage Is Getting Wrong

Most outlets are framing this as "Trump acts on AI" — as if signing a voluntary framework is a meaningful policy move. It isn't.

The NYT calls it "a shift from the hands-off approach." A shift would involve enforcement mechanisms, mandatory timelines, or legal consequences. This order has none of those.

The Daily Signal's coverage, to its credit, accurately describes the order as "scaled-back" and notes the voluntary nature explicitly. The right-leaning press is being more precise here than outlets framing this as a serious regulatory development.

What It Means for You

Powerful AI models that can crack legacy software systems — the kind running power grids, water treatment facilities, and financial infrastructure — are coming to market with no mandatory government review. Companies can ignore this framework entirely with zero legal consequence.

The administration knew the risk. Officials documented it. Then industry made phone calls and the enforcement language disappeared.

Voluntary frameworks are what governments produce when they want to look like they acted without actually doing anything. We've seen this pattern in financial regulation, social media content policy, and data privacy. It never ends well.