A late-night email pinged my inbox from a Commerce staffer; the subject line read, “voluntary submission.” Meta’s lawyers had not signed, and phone calls began to multiply in the West Wing. I watched a soft-power maneuver unfold in real time.
You probably saw the headlines that suggested a regulatory hammer. The executive order President Trump signed on June 2 asks companies to send new models to the Commerce Department’s Center for A.I. Standards and Innovation (CAISI) for a 30-day review while the government finalizes the process by July 31.
Meta hasn’t signed the voluntary pact yet — and that matters
Meta is the most obvious holdout among major AI firms.
While Anthropic, OpenAI, Google, xAI, and Microsoft publicly agreed to submit models to CAISI, Meta has stalled. Their spokesman, Francis Brennan, told reporters the company “shares the administration’s goal of advancing U.S. leadership on robust and secure frontier A.I.” but that they’re “working through the details” and “hope to sign the agreement soon.”
The White House has been personally nudging tech executives; multiple emails and calls have been reported. Commerce officials framed those contacts as routine outreach. Ben Kass, a Commerce Department spokesman, described the push as “the very work CAISI is supposed to be doing” and downplayed any drama.
Does Trump’s AI executive order require companies to comply?
No. The order asks, rather than orders. It invites companies to submit models for a 30-day evaluation window and gives CAISI until July 31 to define the review mechanics. That makes this an exercise in persuasion instead of a legal mandate — a strategic request dressed up as process-shaping.
Anthropic signed on even after a White House intervention — note the irony
Anthropic quietly agreed to the review after its flagship model was taken offline by government action.
That company remains labeled a supply-chain risk by the Pentagon, and Defense Secretary Pete Hegseth has shown no sign of retreating from that designation. Still, Anthropic opted to play along with CAISI’s timeline. Their choice reads as risk management: cooperate publicly while fighting private battles with the Pentagon. The move felt like a velvet glove over an iron fist — soft on the surface but powered by pressure underneath.
Why won’t Meta submit its models for government review?
There are a few plausible reasons. Meta may fear exposing proprietary training data or internal safety processes. It may worry about precedent: once models are handed over, will the government expect access on the next release? And there’s reputational calculus inside boardrooms and legal teams about perceived independence versus political alignment.
Washington chose persuasion over penalties — that’s a tactical decision you should track
The executive order is deliberately gentle with timelines and language.
By asking for voluntary submissions and setting a deadline to define the process, the administration left enforcement options off the table for now. CAISI’s role is framed as review and standards formation, not as a gatekeeper with immediate authority. That ambiguity is strategic: it buys notice and influence without starting a legal fight that could land in federal courts.
Seen another way, the policy is meant to be a beacon; in practice it resembles a lighthouse in a sandstorm — guidance exists, but much of the signal is obscured by corporate caution and competing national-security priorities.
I followed the emails, the calls, and the public statements so you don’t have to take the headlines at face value. The White House asked; Big Tech is answering on its own timetable. So who blinks first, and who gives up leverage to the other side?