Microsoft president Brad Smith says the US is now regulating AI without clear rules. Uncertainty, he warns, is a problem for every industry. He made the case for Fortune on the sidelines of the AI ββfor Good Global Summit.
“What we have right now is regulation without clear or comprehensive rules,” Smith said. “Without laws, businesses cannot organize.”
One blunt instrument
His concern follows two immediate steps taken by the Trump administration. Last month the Commerce Department used export controls to pull Anthropic’s Fable 5 and Mythos 5 models worldwide. It revealed cybersecurity risks. Weeks later, officials pressured OpenAI to delay the public launch of its GPT-5.6 family. Early access to government vetted partners only.
Both curbs have been light. Legend 5 is back online this month, and GPT-5.6 is being released publicly. Smith says Washington was right to act on Fable’s concerns. He says the problem is the tool that got to it.
π for EU tech
The latest talk from the EU tech scene, a story from our genius founder Boris, and some incredible AI art. It’s free, every week, in your inbox. Register now!
“The government has received information that there is an urgent cybersecurity risk. If they receive that information, I think it’s okay,” he said. “But it found it only had one control tool: export controls.” Legal experts recognize that previous controls are models provided by the API. That raises doubts about whether the measure can survive a court challenge.
Rules no one can learn
Critics say the result looks like a licensing regime with no law behind it. The June panel order created a voluntary pre-issuance review, but stopped short of formal approval rules. The government has not yet published who counts as a “trusted partner”, and which models face scrutiny next.
“The government doesn’t have the tools it needs,” said Smith. “Common sense says don’t be heavy-handed, but have enough touch to do what’s needed.”
The problem of trust abroad
The Anthropic episode also lit a fire under autonomous AI. That is pressure on governments to control the models and infrastructure under them. In Europe, one French politician compared the blockade to the blockade of the Strait of Hormuz. Canadian prime minister Mark Carney called it a lesson in dependence on too few suppliers.
Smith thinks people are misreading immigration as an effort to cut off immigration. The goal, he says, was to pull a model for everyone. “They asked Anthropic to take Fable off the market,” he said. “Anthropic said no, so they used the export control lever to make that happen, domestically and internationally.”
Either way, he says the burden now rests with Washington and American companies to prove that their plans are credible. “People will not buy what we sell unless they are sure it will be available,” he said. Salesforce CEO Marc Benioff took the other side at the same event. He felt “good” about the bloc, and said that Europe had misread the security call as hate.
Why is it important
The saga shows how much power the US now holds over AI models that reach the world. It never documented any of those powers. For a company selling AI globally, that’s a serious concern. You can go against the rules. A policy that no one can see is very difficult to plan.
