The FTC wants to treat hidden tuning of AI outputs as deception. Its proposed policy statement, in the Federal Register today, says an AI system that quietly steers answers toward undisclosed objectives, instead of what the user actually asked for, can be a deceptive act under Section 5. The stated way out is disclosure. Notice the mechanism: AI transparency is arriving through ordinary consumer-protection law, not a bespoke AI act. If your product shapes what users see, the question isn't whether that's allowed, it's whether you disclosed it. Comments close 31 July.