FTC Personalized Pricing Closes Friday — The Framework Doesn’t Have an Agent Disclosure Rule

The FTC's proposed enforcement policy on personalized pricing, docket FTC-2026-1057, closes for public comment on September 25 without any provision requiring machine-readable disclosures or rules for AI agents that make purchases on consumers' behalf. Chairman Andrew Ferguson framed the policy around the expectation that a listed price is the same for everyone, requiring companies to disclose that a price is personalized, what drives it, and what data feeds the algorithm, but the framework assumes a human reads the notice. The gap is notable because Meta's Muse launched September 8, SpaceXAI's GrokBot entered beta in August, and Apple's Siri AI reached general availability on September 14, while the 16-state attorney general coalition led by New York's Letitia James and Tennessee's Jonathan Skrmetti filed comments addressing surveillance pricing rather than agent-mediated transactions.

Chairman Andrew Ferguson put the premise plainly: when consumers see a listed price, they expect it to be the same price everyone else sees . The FTC’s proposed enforcement policy https://www.regulations.gov/docket/FTC-2026-1057 on personalized pricing https://forkast.news/algorithmic-pricing/ , which closes for comment on September 25, takes that expectation and builds a disclosure regime around it. Companies using personal data to set individualized prices must tell consumers three things: that the price is personalized, what drives the personalization, and what data types feed the algorithm. The framework assumes someone is there to read those disclosures. That assumption is increasingly wrong. Meta’s Muse launched September 8 and handles purchases on behalf of users. SpaceXAI’s GrokBot entered beta in August. Apple’s Siri AI reached general availability on September 14. Each of these agents makes purchasing decisions for the consumer—selecting products, comparing prices, completing transactions—without a human reading the price tag at the moment of sale. The FTC’s disclosure requirements have no mechanism for this. There is no provision requiring that disclosures be machine-readable, no standard for how an AI agent should interpret or surface personalized pricing information, and no guidance on who bears responsibility when the agent—not the consumer—is the entity encountering the price. The framework is built for a retail environment where humans browse and decide. The market has already moved past that. The structural gap is straightforward. Under the proposed policy, a company using personal data to charge Agent A a different price than Agent B for the same product would need to disclose the personalization to the underlying consumer. But neither Agent A nor Agent B is programmed to process that disclosure. The consumer who delegated the purchasing decision to their AI never sees the notice. The transparency mechanism works on paper and fails in practice. Industry has not addressed this. Public filings in the FTC-2026-1057 docket do not engage with agent-mediated transactions. The 16-state attorney general coalition, led by New York’s Letitia James and Tennessee’s Jonathan Skrmetti, focused its comments on the broader surveillance pricing framework—not on how disclosures function when the purchasing entity is an algorithm. This silence is notable given that three major consumer-facing AI agents launched or entered market in the same month the comment period opened. The FTC’s enforcement posture compounds the problem. The agency is relying on Section 5 of the FTC Act, which prohibits unfair or deceptive practices, to police personalized pricing. But Section 5 is designed to protect consumers who can evaluate representations made to them. When the consumer is not the entity receiving the representation, the legal theory gets thinner. If an agent purchases a product at a personalized price without disclosing the personalization to the consumer, is that a deceptive practice by the seller, a failure by the agent platform, or a gap in the regulatory framework itself? This is not a hypothetical. Prior Forkast coverage has documented the emerging liability architecture around AI agents: the testing-phase liability gap in the EU AI Act When AI Agents Escape Their Sandbox, Who Pays? https://forkast.news/when-ai-agents-escape-their-sandbox-who-pays-europes-product-law-wasnt-built-for-this/ , the principal basis loophole in DHS compliance Post 130578 https://forkast.news/the-principal-basis-loophole-how-dhs-evades-ai-oversight/ , and the broader question of who pays when an agent’s purchase goes wrong. The FTC’s personalized pricing framework sits on top of this unresolved architecture—and the comment period closes in two days. The agency lacks authority to ban personalized pricing outright. Its disclosure framework is the primary oversight tool. If that tool does not account for the fact that AI agents, not humans, are increasingly the ones encountering personalized prices, the framework will regulate a retail environment that no longer exists.