The Trump administration moved Friday, June 27, to assert direct control over access to Anthropic’s most advanced AI model — deciding which companies can use it and which can’t.
It’s a striking intervention. The federal government is now, in effect, acting as a gatekeeper between a private AI company and its potential customers, screening users before they’re allowed in. Anthropic’s model had previously been subject to broader export restrictions; the administration’s action partially lifts that ban while attaching conditions that keep Washington in the loop.
OpenAI’s agreement
OpenAI didn’t wait to be told. The company agreed to let the administration screen users of its own newest model — a concession that hands the federal government an unusual degree of visibility into who is buying access to frontier AI systems.
Neither arrangement has been publicly detailed in full. Which agencies do the screening, what criteria they apply, and whether companies that are denied access have any recourse — none of that’s settled publicly as of Friday.
The moves together represent something that hasn’t existed before in U.S. tech policy: the White House functioning as an approval layer between AI developers and the businesses that want to use their tools. Whether that model holds, gets challenged in court, or gets extended to other AI companies is an open question the administration hasn’t answered.
