When AI Has a Hidden Agenda: The FTC's New Proposal on Output Steerin…

The FTC's 1 July 2026 proposed policy statement targets AI builders and model providers for intentionally steering outputs toward undisclosed objectives (ideological, political, commercial) that differ from user expectations; violations could breach Section 5 of the FTC Act if the steering is not prominently disclosed and matches claims of accuracy or objectivity. If you build or control an AI system, hiding how you steer outputs—even to meet legal or policy goals—is now an FTC violation if you claim objectivity or accuracy. Bury that disclosure in terms of service and the regulator will come for you; make it prominent and you keep the door open to do the steering.