The White House is dictating access to frontier AI models, shifting power from tech giants - CNBC

The Trump administration, freshly installed with a mandate to drain the swamp, has instead decided to drain the AI labs' autonomy. According to *two* anonymous sources — the gold standard for administrative transparency — the White House is now dictating which companies and entities can access frontier AI models. Because handing the keys to the most potent technology since nuclear fission to a political team whose grasp of technical nuance can be summed up as 'the computer says no' is clearly a recipe for competence.

The shift in power from tech giants like OpenAI and Anthropic to the executive branch is being framed as a necessary safety measure. But watch closely: 'safety' in this context means 'control.' The same administration that has spent years gutting regulatory oversight is now asserting unprecedented oversight over AI releases. It's the classic bait-and-switch: 'Trust us, we're from the government, and we're here to help you with your existential risk.' The tech giants, who have been playing fast and loose with safety themselves, now get to blame the government for any future slow-walking of releases. Win-win for everyone except, you know, the planet.

What does 'dictating access' mean in practice? Probably a system of favors, licensing, and political loyalty tests. Korea, China, Iran — who gets the latest model? The answer will depend less on technical readiness and more on diplomatic whimsy. A company that donates heavily to the GOP might get priority access to GPT-7, while a lab with environmentalists on the board might find itself mysteriously delayed. And remember, the administration's track record with technology includes the infamous 'alternative facts' approach. Now they are applying that to model evaluation: 'We have decided that this model is safe because we say so.' Bravo. A new era of AI governance — all the opacity of a black box, with the added thrill of political horse-trading.