Microsoft's Brad Smith: US AI policy lacks transparent rules

Microsoft President Brad Smith used a Fortune interview on the sidelines of the AI for Good Global Summit to say the quiet part out loud: what the US has right now is "regulation without transparent or complete rules," and "without rules, businesses can't plan." Fortune's account frames it as one of the sharpest public breaks yet between a big US vendor and the administration on how frontier AI is being governed.

The episode Smith kept returning to is the Anthropic one. The Commerce Department invoked export-control law to force Anthropic to pull its Fable 5 and Mythos 5 models from the market worldwide, citing a cybersecurity risk, and the models came back online in early July. OpenAI's GPT-5.6 was cleared for a public launch on July 10 after officials had pressed the company to limit access to government-vetted partners. Smith's read is that the underlying security concern was real, but the lever the administration reached for was the wrong shape for the job. Export controls were built for goods crossing borders, not for models delivered over an API, and he put it plainly: the government "doesn't have the tools it needs."

Why that matters if you are not sitting inside a US frontier lab is the international read of the Anthropic action. Smith argued foreign governments interpreted a domestic cybersecurity call as a decision to cut them off specifically, which is exactly the reading that feeds the sovereign-AI scramble European politicians have picked up on. He framed the fix as "technology assurance," meaning predictable rules on both access to markets and reliability of supply.

The honest caveat is that this is one senior executive's framing of a policy his own company benefits from clarifying, and the reporting does not spell out what criteria Commerce is applying, who inside the administration is signing off, or whether the OpenAI delay used the same legal lever as the Anthropic one. What it does establish is that the biggest US software company is now openly describing the current regime as closer to a licensing system without published licenses than to a coherent framework.

The forward-looking piece is who moves next. If Washington will not publish rules, expect Microsoft and its peers to keep pushing for a formal licensing regime that at least trades ambiguity for predictability, and expect European and other non-US buyers to keep hedging with sovereign options while they wait.