Washington's own regulatory dilemma: a federal AI safety board, while Microsoft demands binding global rules

The same day, Microsoft President Brad Smith stated that those voluntary commitments were no longer enough, calling on governments to adopt common, legally binding requirements: independent pre-deployment evaluations, mandatory incident and vulnerability reporting, and shutdown mechanisms for dangerous systems. He cited the United Kingdom as a potential leading player in testing frontier models at international scale.

The timing is not incidental. It marks the moment one of the industry's largest players publicly joins a codification push Congress has only just begun to consider — after two years in which the dominant U.S. industry argument pitted self-regulatory flexibility against what it called the EU AI Act's excessive rigidity.