EPR Editorial Team10 min read
Operation AI Comply is the Federal Trade Commission's enforcement sweep against deceptive claims about artificial intelligence, launched on September 25, 2024 with five actions. Since then the FTC has settled cases against DoNotPay, Evolv, Workado, Growth Cave and Air AI. The standard running through them is plain: a brand must hold competent and reliable evidence for an AI claim before it makes the claim.
Last updated: October 18, 2026.
At a glance
On this page
Operation AI Comply is the name the Federal Trade Commission gave to a September 25, 2024 sweep of five law enforcement actions against companies that used AI claims to deceive consumers or sold AI tools that could be used to deceive them. The FTC announced it under then-Chair Lina M. Khan, who said AI has no exemption from laws already on the books.
The five actions covered DoNotPay, the AI writing tool Rytr, and three online-storefront business opportunity schemes, according to the FTC's launch announcement. The legal basis was Section 5 of the FTC Act, which prohibits unfair or deceptive acts or practices in commerce.
The sweep was a starting point. The cases below, some brought before the sweep and several after it, show how the FTC applies the same deception test to AI claims under two different administrations.
Six FTC actions give brands the clearest picture of what counts as a deceptive AI claim, and the table below summarizes each one before the sections that follow explain what it teaches.
| Case | Key date | What the FTC alleged | Outcome |
|---|---|---|---|
| DoNotPay | Final order, January 16, 2025 | "Robot lawyer" service never tested against human lawyer work | $193,000, notice to 2021 to 2023 subscribers, claim restrictions |
| Rytr | Order December 2024; set aside December 22, 2025 | AI tool could generate fake reviews | Order reopened and vacated, 2-0 vote |
| Evolv Technologies | Stipulated order filed November 26, 2024 | Unsupported claims about AI weapons screening | Claim restrictions, K-12 customers could cancel, no payment, no admission |
| Workado | Final order, August 28, 2025 | "98 percent" accuracy claim for AI Content Detector | Evidence requirement, user notice, compliance reports |
| Growth Cave | Settlement, January 27, 2026 | Business opportunity and credit repair scheme that included AI claims | $48,597,538 judgments, partly suspended; bans |
| Air AI | Settlement announced March 24, 2026 | False earnings and refund claims for AI business tools | $18 million judgment, largely suspended; $50,000 consumer relief; ban |
DoNotPay agreed to pay $193,000 to settle FTC charges that its "robot lawyer" service was never tested against the work of a human lawyer. The FTC said the company also did not hire or retain attorneys to check the quality of its law-related features.
The Commission voted 5-0 on January 16, 2025 to approve the final order, according to the FTC's DoNotPay announcement. The order also bars DoNotPay from advertising that its service performs like a real lawyer unless it has evidence to back that up. The lesson for brands: a comparison to a human professional needs test results behind it before it appears in an ad.
Workado advertised its AI Content Detector as 98 percent accurate, and the FTC alleged that accuracy on general-purpose content was 53 percent. The FTC said the model behind the tool had been tuned for academic writing, not the wider range of text the company marketed it for.
The FTC proposed the order in April 2025 and approved it as final on August 28, 2025, according to law firm Fisher Phillips. The order bars accuracy claims unless they are supported by competent and reliable evidence when they are made. The lesson: a benchmark from one type of data cannot support a claim about every type of data.
Evolv Technologies settled FTC allegations about claims for its Evolv Express weapons screening system in a stipulated order filed on November 26, 2024. The stipulated order bars unsupported claims about weapons detection, false alarm rates, screening speed, labor costs, testing, and the use of artificial intelligence.
The order also gave certain K-12 school customers 60 days to cancel their contracts, and Evolv neither admitted nor denied the allegations. The lesson: a comparison with an existing product, such as a metal detector, is a claim that needs its own evidence, and the remedy can include cancellation rights as well as a claims ban.
Growth Cave settled FTC charges on January 27, 2026 in a case the FTC described as a deceptive business opportunity and credit repair scheme that cost consumers nearly $50 million. The AI claim was one allegation among several. Law firm DLA Piper reports the complaint said the company's GrowthBox "AI software" would automate nearly all of the work of setting up an online course.
The FTC's Growth Cave announcement says the orders ban business opportunity sales and credit repair, and bar misleading claims about earnings, testimonials and the use of artificial intelligence. The judgments total $48,597,538 and are partly suspended because the defendants could not pay in full. The lesson: "AI does it for you" is a performance claim that must match what the software does.
Air AI and its owners agreed on March 24, 2026 to a settlement that bans them from marketing business opportunities and carries an $18 million monetary judgment, largely suspended because they could not pay it. The FTC's August 2025 complaint alleged false claims about earnings, refund guarantees and the performance of Air AI's conversational AI products.
The operators must pay $50,000 to the FTC for consumer relief. The lesson: in this set of cases, the largest judgments, at Growth Cave and Air AI, involved earnings promises rather than product accuracy claims, so income and refund statements tied to an AI product carry the highest financial exposure.
The FTC's December 22, 2025 decision to set aside its Rytr order narrowed one theory of liability, not enforcement of AI claims as a whole. Chairman Andrew N. Ferguson and Commissioner Mark Meador voted 2-0 to reopen and vacate the 2024 consent order, which had barred Rytr from offering any AI service that generates reviews or testimonials.
The two commissioners concluded the original complaint did not support a Section 5 violation and that the order burdened AI innovation, citing the July 2025 America's AI Action Plan, according to DLA Piper and Broadband Breakfast. The practical result is that a tool is no longer presumed deceptive only because customers could misuse it.
Companies that misstate what their AI does remain exposed. Growth Cave and Air AI were both resolved after the Rytr decision.
The FTC's proposed policy statement, released July 1, 2026, says that steering an AI system's outputs toward objectives the user did not request and was not told about can deceive consumers under Section 5. It is titled "Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems," and the full proposed policy statement is posted in the federal docket.
The proposal follows Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence," signed December 11, 2025, which directed the FTC to address state laws that require changes to truthful AI outputs. The statement names Colorado's revised Artificial Intelligence Act as an example and says that complying with a state law does not excuse undisclosed steering. It also says a company can reduce deception risk by truthfully disclosing what its model prioritizes.
Law firm Hall Render notes that hallucinations alone are not the core conduct targeted, though misrepresenting model reliability can still raise deception issues. The comment period closed July 31, 2026. A proposed policy statement is not a binding regulation, and the most recent coverage reviewed for this update, dated August 2026, still described it as proposed.
Five patterns recur across the FTC's AI cases, and each one maps to a kind of claim a brand can test before publishing it.
Fake and AI-generated reviews, the subject of the original Rytr case, fall under a separate FTC rule explained in Everything-PR's guide to the FTC fake review rule.
A brand should build a written evidence file for every AI claim before the claim runs, because the FTC orders in these cases ask whether the evidence existed at the time the claim was made. The steps below turn that standard into a checklist.
The through-line in every case is timing. The FTC asks what a company knew when the ad ran, so the evidence file has to exist before launch, not after a Civil Investigative Demand arrives. For related disclosure questions, see the platform rules for AI content disclosure.
Operation AI Comply is the FTC's September 25, 2024 enforcement sweep against deceptive AI claims and AI-enabled schemes, announced with five law enforcement actions. The FTC has continued to bring and resolve AI-related cases since then.
Yes, the FTC's orders require evidence behind AI claims. The Workado order requires competent and reliable evidence at the time a claim is made, the Evolv order defines that evidence as objective tests by qualified professionals, and the DoNotPay order bars lawyer-level claims without sufficient evidence.
No, the FTC continued enforcement after December 22, 2025. It announced the Growth Cave settlement on January 27, 2026, the Air AI settlement on March 24, 2026, and a proposed AI accuracy policy statement on July 1, 2026.
Penalties range from conduct orders to large judgments. DoNotPay paid $193,000, Growth Cave's judgments total $48,597,538 and are partly suspended, and Air AI's $18 million judgment is largely suspended with $50,000 going to consumer relief. The Evolv order carried no payment, and the Workado order centered on evidence requirements, user notice and compliance reporting.
Not yet, because it is a proposal and not a binding regulation. It targets undisclosed steering of AI outputs, and the existing Section 5 deception test continues to govern claims about what an AI product can do.
Yes, if the brand makes claims about AI in its own marketing. The orders turn on whether a claim is true and supported, so a brand that advertises a product as AI-powered carries the same evidence burden as an AI vendor.
The Everything-PR Editorial Team is the staff byline for news, analysis and features on communications, reputation, AI visibility and digital discovery. Everything-PR has published since 2009. AI tools assist with research and drafting, and every article is reviewed by a human editor before publication. Coverage follows the Editorial Policy, and substantive corrections are noted on the article under the Corrections Policy.
FleishmanHillard's Intel program is the canonical modern tech-launch playbook. Inside the disciplines that make a major hardware launch work — reviewer relationships, benchmark methodology, analyst relations, retailer editorial, trade press — and what brands beyond Intel are worth studying.
OpenAI and Anthropic built the foundational model layer that sits underneath every major AI Communications surface. ChatGPT and Claude define what gets cited, what gets generated, and how brand reputation gets distributed in the answer-engine era.
The 2026 Clio Awards named 20 Grand Clio winners. See every winner, AXA's THREE WORDS as the Public Relations Grand, jury chair Josy Paul, and 2027 entry deadlines starting October 23, 2026.
Compare the best reputation management companies of 2026. Everything-PR ranks 30 firms across crisis, ORM, software and PR, scored on the same seven criteria.
PR news this week: O'Dwyer's 2026 rankings, PR agency M&A, new offices and senior communications hires. Updated every Thursday, with the September 14 issue kept below.
Yankees marketing and public relations explained: a $8.5B brand, HOPE Week, the YES Network, the 2025 beard rule change and Aaron Judge's captaincy story.
EPR publishes the data every week.
Free. Weekly. Unsubscribe anytime.
