An FTC sweep tested AI marketing claims, not companion design

The design

On 25 September 2024 the FTC announced Operation AI Comply, a sweep against five operations it said used AI hype, or sold AI tools usable, to deceive consumers. The release bundled a settlement with DoNotPay over an 'AI lawyer' service, a fake-review generator called Rytr, and three online-storefront schemes — Ascend Ecom, Ecommerce Empire Builders, and FBA Machine/Passive Scaling — alleged to promise AI-powered passive income while delivering little. The FTC's own case page for DoNotPay shows the company agreed to pay $193,000 and warn past subscribers, with a final order issued in January 2025.

What the evidence says

None of the five cases names a companion or relationship chatbot; the alleged deception concerned legal-service claims, review authenticity, and guaranteed income, not emotional support. What the announcement establishes is an enforcement posture: Chair Lina Khan is quoted saying there is 'no AI exemption from the laws on the books,' and the release ties the sweep to earlier cases, including one against a messaging app's AI-moderation claims. Reading a companionship-specific finding into this sweep would extrapolate beyond the FTC's own case list; the narrower, supportable reading is that an AI product's marketing claim, including a claim of emotional benefit, faces ordinary deception law regardless of the technology.

What it asks of people

For companies marketing AI products, the sweep asks for evidence behind performance claims before they are made; the DoNotPay complaint centers on the company's failure to test whether its chatbot matched a human lawyer's output. For a reader evaluating a companion app's marketing, it suggests one question: does a claim about accuracy, memory, or emotional benefit rest on disclosed testing, or is it, like these five cases, an unverified promise attached to a language model.

Privacy and safeguards

The sweep is a set of enforcement actions and orders, not a rulemaking; it creates no disclosure duty specific to companion or emotionally engaging AI. Its orders bar the named companies from repeating the claims at issue and, for Rytr, from selling review-generation services at all, but say nothing about data retention, minors, or relationship-design features a companion app might use. The gap between an enforcement sweep on marketing fraud and a safeguard on product design is worth naming, since summaries easily conflate the two.

• Does the company behind a companion app disclose testing behind claims about memory accuracy, emotional benefit, or safety?
• Has that company, or a similar one, appeared in an FTC case list for deceptive AI marketing?
• Does 'no AI exemption from the law' reach product-design harms, or only marketing claims about capabilities?

Operation AI Comply is most useful as a precedent for how the FTC treats AI marketing claims generally: ordinary claims subject to ordinary proof. It is not, on its own record, an action against companion products, and describing it that way overstates what the agency's case list shows.