FTC proposes AI accuracy and manipulation policy statement

The Federal Trade Commission has proposed a policy statement that would treat some forms of undisclosed ideological steering in artificial intelligence systems as potentially deceptive under § 5 of the FTC Act, setting up a comment period that runs through July 31, 2026.

Published on July 1, the proposal focuses on whether AI companies may be violating federal law if they direct system outputs toward undisclosed ideological objectives instead of the objectives consumers request or reasonably expect. The FTC said the issue can arise whether a company does so on its own or in response to a state law requirement.

Focus on consumer expectations

According to the proposed statement, AI companies have made explicit and implicit representations that their systems are built to deliver the best, most accurate, and most faithful output possible within technological and resource constraints. The agency’s view is that those representations shape what users reasonably expect from AI tools.

The statement also says consumers accept AI outputs without independent fact-checking more than 90% of the time. On that basis, the FTC argues that users reasonably expect AI systems not to be secretly designed to pursue undisclosed objectives that distort outputs.

Using its three-part deception framework, the FTC said conduct is deceptive if there is a representation, omission, or practice likely to mislead a reasonable consumer in a material way. The proposal concludes that intentional ideological steering could fall within that standard.

The agency drew a line between deliberate design choices and technical errors, saying AI "hallucinations" result from technological and resource limitations rather than design decisions and, by themselves, do not raise § 5 concerns.

sbb-itb-e766981

Executive order and state-law conflict

The proposal was issued pursuant to Executive Order 14365, signed by President Trump on December 11, 2025. The order directed the FTC to clarify how § 5 of the FTC Act applies to AI models and to address how state laws requiring alterations to accurate AI outputs can conflict with federal law.

The statement also reflects what the source article described as the Administration’s broader goal of establishing a national AI regulatory framework that preempts state-by-state regulation.

A major element of the proposal is its treatment of state AI laws. The FTC points to Colorado’s Artificial Intelligence Act as an example of a law that may pressure AI companies to suppress output accuracy to avoid disparate impact liability. The statement concludes that such a state law is impliedly preempted to the extent it conflicts with the federal regulatory scheme established by § 5 of the FTC Act.

The FTC’s position is that a state law that effectively requires an AI company to deceive consumers conflicts with § 5’s express purpose of protecting consumers from deceptive conduct.

Disclosure safe harbor

The proposal says companies can avoid § 5 liability if they make clear, conspicuous, and adequate disclosures that a system is designed to prioritize certain objectives over what users request or would otherwise expect. But it sets what the source article described as a high bar for those disclosures.

A disclaimer buried in terms of service would not be enough, according to the statement. Instead, disclosures would need to be prominent and persistent enough to change the reasonable consumer expectations created by a company’s marketing and by AI’s role as a problem-solving tool. The proposal adds that the more a disclosure conflicts with expectations created elsewhere, the more prominent and persistent it must be.

Implications for companies

The proposed statement has practical implications for AI developers and companies deploying AI-powered products and services, according to the source article.

On product design and marketing, companies that present AI systems as accurate, neutral, or objective problem-solving tools are urged to assess whether actual system design and outputs match those representations and whether any intentional output constraints are adequately disclosed.

On state law compliance, companies subject to state AI laws, including Colorado’s revised Artificial Intelligence Act, may need to evaluate whether compliance steps could be characterized as intentional output suppression that conflicts with consumers’ reasonable expectations and with § 5.

On disclosure practices, companies that configure systems to prioritize objectives beyond what users would otherwise expect may need to review whether disclosures in terms of service, product descriptions, and onboarding flows are prominent enough to meet the standard described in the proposal.

Public comments on the FTC’s proposed policy statement are due by July 31, 2026.

Read the source