Who Has Authority to Declare AI Safe?

A control gap exists between what AI can do and what evidence supports trusting it to do, according to analyst Dave Vellante. The September 2026 White House accord includes independent board-level oversight but creates no public authority with mandatory access or enforcement powers. Appian CEO Matt Calkins argues government should require AI companies to prevent damage now, drawing parallels to banking and nuclear safety regulation. Vellante contends that independent experts must examine AI evidence and that an authority with real enforcement power is needed. As of now, the framework remains fuzzy, with tech leaders assigned only moral authority.