White House backs lighter AI approach

Artificial intelligence policy is suddenly looking less like a new federal rulebook and more like a handshake with guardrails. That is the real news here. After Kevin Hassett suggested the White House was considering something like FDA-style pre-release testing for advanced AI models, other senior officials rushed to cool that down and say the administration wants “partnership” with companies, not broad new regulation. ### What actually happened? On May 7, Kevin Hassett, who runs the National Economic Council, said the administration was studying an executive order that could make future AI systems go through a safety process before release — “just like an FDA drug.” That landed as a big shift because it sounded like the White House might force companies such as OpenAI, Anthropic, and Google into federal review before launching top-tier models. (politico.com) ### Why did the White House walk it back? Because the reaction was immediate. Industry people heard “FDA for AI” and saw a world where model launches get delayed, startups get boxed out, and Washington becomes the gatekeeper for who ships what. By the next day, a senior White House official was saying Hassett’s comments were taken “out of context a little bit” and stressing that only a minority inside the administration wants hard regulation. Susie Wiles piled on, saying the government is “not in the business of picking winners and losers.” (politico.com) ### So is federal vetting dead? Not exactly. The catch is that the White House seems to want the option without fully owning the label. The Politico report says officials are still discussing a vetting system for advanced models, especially because newer systems can expose software flaws and raise fears about cyberattacks or even bioweapon-related misuse. So the administration is not saying “no oversight.” It is saying “don’t call this a heavy regulatory regime” — at least not yet. (politico.com) ### Why is the administration leaning this way? Because this has been the direction of travel for months. Trump’s December 11, 2025 executive order framed AI as a strategic race and said U.S. companies must be free to innovate without “cumbersome regulation.” It also pushed for a single, minimally burdensome national approach instead of a patchwork of state laws. Then in March 2026, the White House published legislative recommendations that again favored a federal framework with targeted protections but lighter compliance burdens overall. (politico.com) ### What does “partnership” really mean? Basically, it means the government wants companies to do more of the safety work themselves. Firms may still be expected to share information, consult with national security agencies, and show they have internal controls before releasing very capable models. But the burden shifts away from a formal licensing-style system and toward voluntary cooperation, self-governance, and selective government pressure behind the scenes. That is friendlier for incumbents and much easier to sell politically. (whitehouse.gov) ### Who benefits from that? Startups probably get the clearest upside. A full pre-clearance system would hit smaller labs hardest because they have less legal staff, less cash, and less time to wait. The White House’s own December order explicitly argued that fragmented state regulation is especially hard on startups. A lighter federal touch lowers those barriers — but it also means smaller companies have to build their own safety processes without much regulatory certainty. (politico.com) ### What is still unresolved? A lot. Who decides which models are dangerous enough for extra scrutiny? What counts as adequate testing? What happens if a company ignores “partnership” and ships anyway? And if Washington preempts states while avoiding strong federal rules, the result could be less oversight overall, not smarter oversight. That is the part everyone is now squinting at. (whitehouse.gov) ### Bottom line? The White House is trying to keep two things at once — fast AI deployment and some ability to intervene when models look risky. But those goals pull against each other. Right now, the administration is clearly choosing the lighter-touch story first, and leaving the enforcement details for later. (politico.com)