FTC proposes treating hidden ideological tuning of AI outputs as deceptive

The US Federal Trade Commission voted 2-0 on July 1 to propose a policy statement arguing that AI companies which distort model outputs for undisclosed “ideological objectives” can be committing deception under Section 5 of the FTC Act. The agency opened a public comment period running through July 31.

The move is significant because it applies the FTC’s core consumer-protection authority to how AI labs tune their models, an area regulators have largely left alone. The statement, titled “Suppression of Accuracy in Artificial Intelligence Systems,” contends that users treat AI answers as reliable: it cites that consumers accept AI outputs without independent fact-checking more than 90 percent of the time and reasonably expect a system to produce its most accurate output. On that logic, quietly steering a model away from accuracy to serve an undisclosed viewpoint could deceive them.

The proposal was issued pursuant to Executive Order 14365, signed by President Trump on December 11, 2025, which directed the agency to clarify how Section 5 applies to AI models. The FTC’s request for public comment is a proposed statement, not a rule or an enforcement action, and it does not name any company. The full policy statement is available on the FTC’s legal-library site.

The practical reach is untested. A policy statement signals how the commission reads the law rather than creating a new legal standard, and any deception case would still have to prove a specific, material and undisclosed distortion, a high bar given how opaque model-tuning decisions are. The framing also raises a line-drawing problem: separating safety and bias-mitigation tuning from “ideological” manipulation is exactly what the comment period invites the industry to contest.

The deadline is the near-term marker: watch whether the FTC finalizes the statement after July 31 or softens the “ideological” framing under industry and legal pushback.