FTC Targets AI 'Suppression of Accuracy' as Consumer Deception
A new proposal would penalize AI developers who deliberately steer model outputs away from accuracy without disclosing those goals to users.
The Federal Trade Commission has proposed a new policy to target AI companies that intentionally steer their systems' answers away from accuracy toward other undisclosed goals. Published on July 1, 2026, the proposal frames this "suppression of accuracy" as a form of consumer deception under Section 5 of the FTC Act.
According to the proposal, the FTC intends to penalize companies that configure their AI to favor specific goals over accuracy on purpose without informing the user. The agency makes a critical distinction between this intentional "output steering" and technical "hallucinations." While hallucinations are viewed as results of current technological limitations, the deliberate modification of outputs to serve a hidden agenda is treated as a deceptive trade practice. This includes scenarios where a company might modify outputs to comply with specific state laws, such as the Colorado AI Act, without disclosing those adjustments to the end user.
The Regulatory Landscape
This move introduces a federal layer of oversight that differs from existing global and domestic frameworks. While the European Union's AI Act mandates strict accuracy standards for high-risk systems, and various U.S. state laws focus on preventing discrimination in sectors like lending or hiring, the FTC is focusing specifically on the deceptive nature of the interaction. By utilizing Section 5 of the FTC Act, the commission is positioning undisclosed steering not as a technical failure, but as a breach of trust between the provider and the consumer.
Industry Implications
If adopted, this policy would force AI developers to be transparent about the "guardrails" and alignment goals programmed into their models. It creates significant legal risk for companies attempting to align AI outputs with corporate or political objectives if those adjustments contradict the system's presented accuracy. Developers may find that simply complying with state-level regulations is insufficient to satisfy federal requirements if the mechanisms used to achieve that compliance are hidden from the user.
What to Watch
As the FTC seeks comment on the proposed policy statement, the industry will be watching how the agency intends to prove intent. Because the policy targets "intentional design" rather than technical error, internal company records and design documents are expected to become central to future enforcement cases. It remains to be seen how the FTC will distinguish between a model's inherent bias and a developer's deliberate steering in a court of law.