White House officials may expand voluntary AI safety reviews to powerful open models, raising new questions about security, competition, and compliance.
