FTC Act Section 5 (Unfair or Deceptive Practices) — Regulatory Reference

Section 5 of the FTC Act (15 U.S.C. § 45) prohibits unfair or deceptive acts or practices in commerce. For retail AI deployments, this means AI agents used in customer-facing interactions — pricing, personalization, recommendations, chatbots — must not produce outputs that mislead consumers or cause substantial harm they cannot reasonably avoid. The FTC's 2023 guidance on AI claims confirmed that Section 5 applies to AI-generated content and automated decisions. Retailers must be able to demonstrate what data an AI agent consumed, what decision it made, and why — which requires a governance layer that captures pre-retrieval enforcement and a tamper-evident record of every agent decision.

Section 5 applies whenever an AI agent's output could deceive or materially harm a consumer in connection with a commercial transaction. In retail, this includes dynamic pricing models that apply differential rates without disclosure, recommendation engines that suppress options based on protected or sensitive attributes, and chatbots that make product or financing claims the underlying data does not support. The Section 5(n) unfairness test — substantial injury, not reasonably avoidable, not outweighed by countervailing benefits — is the operative standard. The FTC has cited this test in enforcement actions involving algorithmic decision-making since 2024, making it directly relevant to any retailer deploying AI agents in customer journeys.

The FTC can seek civil penalties of up to $51,744 per violation per day for repeat or knowing violations under Section 5. First-time violations typically result in consent orders requiring corrective action and ongoing compliance reporting — but subsequent violations of a consent order trigger per-day monetary penalties. The FTC has broad investigatory authority: it can demand internal records, audit logs, model documentation, and evidence of consumer harm. In 2024 settlements involving biased and deceptive AI, the FTC required companies to implement AI governance programs, maintain documentation of algorithmic inputs, and submit to third-party audits. Retailers without a documented audit trail of AI agent decisions face significant exposure if the FTC opens an inquiry.

AutoPIL enforces access policy before sensitive data enters an AI agent's context window — preventing agents from consuming data categories that could generate deceptive personalization or unfair differential treatment. The agent registry distinguishes customer-facing agents from internal ones, so governance controls can be scoped to the interactions that carry FTC exposure. Every enforcement decision is written to a tamper-evident audit chain, giving compliance teams a complete, unalterable record of what data each agent accessed and what policy governed that access. Two specific policy configurations — RET-FTC5-UD-001 (Deceptive AI Output Boundary) and RET-FTC5-UF-001 (Unfair Personalization Audit) — map directly to the Section 5 unfairness and deception tests and can be adapted to a retailer's data environment.

Section 5(n) defines an unfair act or practice as one that causes or is likely to cause substantial injury to consumers, that consumers cannot reasonably avoid, and whose harm is not outweighed by countervailing benefits. Applied to AI in retail, an agent that uses consumer financial stress signals or browsing history to inflate prices on essential goods — without the consumer's knowledge or any opt-out mechanism — meets all three prongs. The FTC has applied this standard to algorithmic practices since its 2022 policy statement on unfair methods of competition. For retailers, satisfying the 5(n) defense requires evidence that the AI system had defined data access boundaries, that those boundaries were enforced at runtime, and that a record exists showing the system operated within policy.