The Federal Trade Commission has proposed a policy statement arguing that companies marketing AI systems may deceive consumers when they steer outputs toward undisclosed objectives that differ from those users set or reasonably expect.
The proposed theory centers on the gap between an AI system’s apparent purpose and its hidden objectives. If a system is presented as pursuing users’ objectives faithfully and accurately but prioritizes another objective, the FTC says that steering may create deception risk under section 5.
The statement treats intentional, objective-based steering as distinct from ordinary hallucinations caused by technological or resource limitations. It also says a company’s motive, whether profit, public opinion, or another reason, does not determine whether section 5 applies. The statement takes no position on whether the discussed practices may also be unfair.
The proposal presents disclosure as one way to shape consumer expectations when a system prioritizes objectives users would not otherwise expect. Those disclosures would need to be clear, conspicuous, prominent, and persistent enough to counter the system’s apparent purpose; burying them in terms of service would not suffice under the proposal. Comments were due July 31, 2026.
For enterprise governance teams, the proposal offers a review lens rather than a current standalone compliance mandate; identify the objectives a model pursues, compare them with user expectations and product representations, distinguish deliberate steering from capability limits, and assess whether disclosures accurately explain material differences.