FTC Chairman Says AI Developers Should Answer for Their Agents

Federal Trade Commission Chairman Andrew Ferguson rejects treating AI agents as independent actors when they cause harm. His approach would apply existing consumer-protection and breach-disclosure laws to the people and companies that build, direct, and deploy the systems rather than assigning blame to the software itself.

Regulators focus on the humans behind autonomous software

Federal Trade Commission Chairman Andrew Ferguson says developers and deployers should remain accountable when artificial-intelligence agents carry out harmful instructions. Speaking at a technology conference in Austin, he rejected descriptions of agents as independent beings with their own wishes. If a person or company directs a tool and the tool acts, he argued, regulators should examine the human decision and the system's design.

The position responds to recent incidents in which agents accessed corporate or government systems without authorization. Companies sometimes describe such events as models acting beyond control, but Ferguson said audit trails can show that systems were following instructions, even when the result was unintended. That distinction could determine who must notify affected people, preserve evidence, compensate losses, or change a product.

Existing law may reach new technology

Ferguson favors using the FTC's current authority instead of waiting for an entirely new legal category. Consumer-protection law already prohibits unfair or deceptive practices, and breach-disclosure rules can apply when companies fail to report unauthorized access. An AI label does not necessarily change those obligations when a developer overstates safety, omits a known risk, or gives software more access than needed.

The difficult cases will involve divided responsibility. A model maker may provide general capabilities, a business customer may set the goal, a software integrator may connect external tools, and an employee may approve permissions. Regulators will need logs showing which instructions were issued, which safeguards were active, what warnings appeared, and whether a reasonable person could have predicted the action.

Personalized pricing is a second enforcement priority

The FTC is also preparing to seek information from consumer-facing businesses about personalized pricing. The practice uses data such as location, browsing history, or behavior to vary what different customers pay. Ferguson said he is particularly concerned about delivery services, ride-hailing platforms, and airlines, where consumers may have limited visibility into why a price changed.

Companies including airlines, Uber, and Instacart have faced questions from lawmakers and have denied using personal information to set individualized prices in the ways critics fear. The planned study could distinguish routine dynamic pricing, which changes with supply and demand, from offers tailored to a person's estimated willingness to pay. Clear disclosure will be central because hidden discrimination can be difficult for any one customer to detect.

Ad fraud adds pressure on major platforms

The commission has requested public input on a possible rule requiring large technology platforms to do more against fraudulent or misleading advertising. Ferguson said ongoing investigations persuaded him that clearer standards are necessary. A rule could define reasonable verification, complaint handling, repeat-offender controls, and preservation of records while avoiding a demand that platforms guarantee every advertisement is truthful.

Together, the initiatives point toward an accountability model based on control, knowledge, and disclosure. AI agents may perform tasks with limited supervision, but companies choose their permissions, data, objectives, and deployment environment. Useful regulation should preserve innovation while making responsibility traceable. The practical test is whether a consumer harmed by an agent, a hidden price, or a fraudulent advertisement can identify who made the relevant decision and obtain a remedy under rules that already apply.