Microsoft’s Brad Smith: the US is regulating AI with rules nobody can read

Microsoft president Brad Smith says the US is now regulating AI without a clear set of rules. The uncertainty, he warns, is a problem for the whole industry. He made the case to Fortune on the sidelines of the AI for Good Global Summit.

“What we really have right now is regulation without transparent or complete rules,” Smith said. “Without rules, businesses can’t plan.”

One blunt tool

His concern follows two abrupt moves by the Trump administration. Last month the Commerce Department used export-control law to pull Anthropic’s Fable 5 and Mythos 5 models worldwide. It cited a cybersecurity risk. Weeks later, officials pressed OpenAI to delay the public launch of its GPT-5.6 family. Early access went to government-vetted partners only.

Both curbs have since eased. Fable 5 came back online this month, and GPT-5.6 is rolling out publicly. Smith says Washington was right to act on the Fable concern. The problem, he argues, is the tool it reached for.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!

“The government got information that there was an urgent cybersecurity risk. When it gets that information, I think it’s right to act,” he said. “But it found it had only one regulatory tool: an export control.” Legal experts note those controls predate API-served models. That raises doubts about whether the move could survive a court challenge.

Rules nobody can read

Critics say the result looks like a licensing regime with no legislation behind it. A June executive order set up a voluntary pre-release review, but stopped short of formal approval rules. The government has not published who counts as a “trusted partner”, nor which models face vetting next.

“The government doesn’t have the tools it needs,” Smith said. “Common sense says don’t be heavy-handed, but have enough of a touch to do what’s needed.”

The trust problem abroad

The Anthropic episode also lit a fire under sovereign AI. That is the push by governments to control the models and the infrastructure beneath them. In Europe, one French politician likened the shutdown to a blockade of the Strait of Hormuz. Canadian prime minister Mark Carney called it a lesson in leaning on too few providers.

Smith thinks people misread the export move as an effort to cut off foreigners. The intent, he says, was to pull the model for everyone. “They asked Anthropic to take Fable off the market,” he said. “Anthropic said it would not, so they used an export-control lever to make it happen, at home and abroad.”

Either way, he says the burden now sits with Washington and US firms to prove their systems are dependable. “People will not buy what we sell unless they’re confident of certainty of supply,” he said. Salesforce chief Marc Benioff took the other side at the same event. He felt “good” about the block, and said Europe had misread a security call as hostility.

Why it matters

The saga shows how much power the US now holds over which AI models reach the world. It has barely written any of that power down. For a company that sells AI globally, that is the deeper worry. You can argue with rules. A policy nobody can see is far harder to plan around.