The Federal Trade Commission published a proposed policy statement July 1 targeting AI companies that steer their systems toward undisclosed ideological objectives. The FTC is taking the position that such manipulation may violate Section 5 of the FTC Act, which prohibits deceptive or unfair practices.
The public has until July 31 to submit comments. The policy statement is a significant federal move to regulate AI output integrity. It’s also preemptive, attempting to override state laws like Colorado’s Artificial Intelligence Act that the FTC sees as forcing companies to suppress accuracy.
What the FTC is Saying
The core argument: AI companies have made explicit and implicit representations that their systems produce the best, most accurate, most faithful output possible. Consumers reasonably expect this. If a company secretly designs its AI to pursue undisclosed objectives that distort those outputs, that’s deceptive.
The statement singles out Colorado’s AI law as an example of state regulation that might pressure companies to suppress output accuracy. The FTC is essentially saying: federal law preempts state law here. Companies can’t use state rules as cover for manipulation.
How Companies Can Comply
The path is clear: disclose. If an AI company is designed to prioritize certain objectives over what users request, make that clear and conspicuous in the user interface and terms of service. Transparency defeats the deception claim.
This opens a door: companies could say, “Our AI is designed to refuse certain queries” or “Our AI prioritizes certain values in outputs.” As long as it’s disclosed, the FTC’s framework allows it.
The Bigger Picture
This is federal regulation of AI taking a specific angle: consumer protection through accuracy and transparency. It’s not a ban on alignment techniques or value-loading. It’s a requirement to be honest about what the system does.
The comment period through July 31 will reveal how AI companies plan to respond. Expect pushback from some, compliance statements from others.
The FTC is drawing a line: you can build AI systems however you want, but you can’t lie about what they do.
