Brad Smith calls for mandatory AI incident reports. Microsoft's president set out his proposals in a BBC interview.
Companies should document incidents and notify victims when uncontrolled AI agents breach their systems, Smith says.
He proposes government evaluators before new AI releases and technical emergency brakes. Voluntary self-regulation is no longer sufficient, he argues.
These are proposals, not adopted rules. The BBC published the interview on October 11, 2026; its recording date remains unknown.