Opinion | Who Enforces the AI Rules?

Peggy Noonan is right to take artificial intelligence risks seriously (“Trump Is in Denial About AI’s Danger,” Declarations, Sept. 26). The reported incident in which AI agents breached computer systems deserves scrutiny. But her call for a cooperative safety regime echoes other flawed approaches to policy by skipping the question that determines whether such a regime would work: What behavior would it prohibit, how would violations be detected, and who would enforce the rules?

A promise to make AI “safe” can’t be verified. Specific safeguards can be tested: limits on an agent’s access, independent security evaluations, reporting of serious incidents and the ability to shut down a system that acts outside its authority. Those measures would address the failure Ms. Noonan describes more directly than any toothless international declaration.

Copyright ©2026 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8