FTC warns AI accuracy changes may be deceptive

The FTC has issued a proposed policy statement warning that AI developers may violate Section 5 of the FTC Act if they steer outputs away from users’ expected objectives without adequate disclosure. The statement says companies that market their systems as producing faithful and accurate results could deceive consumers when hidden design choices undermine those expectations.

The proposal applies the FTC’s established deception framework to AI systems and emphasizes that disclaimers must be clear, conspicuous and prominent. The agency distinguishes deliberate accuracy suppression from hallucinations caused by technological limits, though misrepresenting the likelihood of hallucinations could still raise deception concerns.

The policy also challenges the idea that state law compliance can justify undisclosed changes to AI behavior. Colorado’s revised Artificial Intelligence Act, adopted in May 2026, is cited as a law that could push companies to adjust outputs to avoid discriminatory outcomes. The FTC takes the position that state law is not a defense if those adjustments conflict with federal consumer protection rules.

Financial services firms and vendors face potential scrutiny as AI becomes more common in lending, underwriting, customer service and financial advice. The public comment period closes July 31, 2026, giving companies limited time to assess whether governance, bias testing and output adjustments are adequately disclosed.