FTC Operation AI Comply: Cases, the Rytr Reversal, and the 2026 Playbook

Last reviewed: April 29, 2026

The FTC has become the most active US federal agency enforcing AI-related violations, building a record of 8 cases through Operation AI Comply plus a parallel track of pre-Operation algorithmic disgorgement actions — all without ever drafting an AI-specific rule. The agency relies on Section 5 of the FTC Act — the same century-old “unfair or deceptive acts or practices” authority it uses for everything from telemarketing fraud to data privacy. In December 2025, Chairman Andrew Ferguson set aside the Rytr consent order on the grounds that it “unduly burdens AI innovation,” signaling a recalibration. But the broader enforcement program continues. This article catalogs every public case, extracts the patterns, and translates them into a six-step compliance playbook.

Key Takeaways

8 Operation AI Comply cases: 5 original actions in September 2024 plus 3 follow-on cases (Click Profit, Workado, Air AI) through March 2026. Total consumer losses across the cases exceed $73 million.
The Rytr reversal is narrow but consequential: the December 22, 2025 set-aside limits the FTC’s “means and instrumentalities” theory of liability. It does not signal a retreat from enforcing genuinely deceptive AI claims.
Algorithmic disgorgement remains the FTC’s most powerful remedy: the agency has ordered model deletion in four cases (Everalbum, WW/Kurbo, Rite Aid, Amazon Alexa). The remedy is unchanged under the new administration.
The March 11, 2026 FTC AI Policy Statement reframes Section 5 around AI: warning phase through 2026, full enforcement (penalties up to $53,088 per violation) starting 2027.
The pattern is consistent: the FTC targets demonstrable consumer harm tied to false claims, not the existence of AI tools themselves. Compliance comes down to substantiating what you advertise.

What is the FTC’s authority to regulate AI?

The FTC enforces AI under existing law. There is no AI-specific federal statute. Section 5(a) of the FTC Act prohibits “unfair or deceptive acts or practices in or affecting commerce” — language that has been on the books since 1914. The Magnuson-Moss Warranty Act, the Children’s Online Privacy Protection Act (COPPA), the Equal Credit Opportunity Act (ECOA), and the Fair Credit Reporting Act (FCRA) each give the FTC additional jurisdiction in specific contexts.

The agency’s view, articulated in its March 11, 2026 AI Policy Statement, is that Section 5 already covers AI’s two main consumer harms: deception (false claims about what an AI system does) and unfairness (algorithmic outcomes that injure consumers without offsetting benefits). The Policy Statement explicitly avoids new rulemaking. It interprets existing authority instead.

This matters because Section 5 enforcement does not require notice-and-comment rulemaking, congressional action, or sector-specific authority. The FTC can bring a case as soon as it has reason to believe an AI claim is unsubstantiated. That speed is unusual in US federal enforcement.

Chairman Ferguson, who took the chair in January 2025, has emphasized this distinction repeatedly. In his December 2025 Rytr set-aside order, he wrote that “consumers benefit from the invention and availability of new tools” and that the FTC should police AI providers using “well-established deception authority — such as when companies misrepresent their AI capabilities.”

What cases has the FTC brought under Operation AI Comply?

Operation AI Comply launched on September 25, 2024 with five simultaneous actions. The FTC has brought three additional AI-related cases since (Click Profit in March 2025, Workado in April 2025, Air AI in August 2025); the Rytr order from the original sweep was reopened and set aside in December 2025. The cases break into two categories: AI-powered scams (false earnings claims dressed up as AI) and false AI capability claims (products that do not perform as advertised).

Case	Filed	Outcome	Consumer Loss	Theory
DoNotPay	Sept 2024	$193K consent order; banned from “robot lawyer” claims (Feb 2025 final)	Not quantified	False AI capability claims
Ascend Ecom	Sept 2024	TRO; permanent ban on business opportunities (2025)	$25M+	AI-powered earnings scheme
Ecommerce Empire Builders	Sept 2024	Permanent ban; asset turnover (May 2025)	Not quantified	AI-powered earnings scheme
FBA Machine	Sept 2024	Permanent ban; asset seizure (July 2025)	$15M+	AI-powered earnings scheme
Rytr	Sept 2024	Consent order set aside Dec 22, 2025 (2-0)	Not quantified	Means and instrumentalities
Click Profit	March 2025	$13.6M judgment (Aug 2025); permanent industry ban	$14M+	AI-powered earnings scheme
Workado	April 2025	Order requires substantiation (Aug 2025 final, Docket C-4822)	Not quantified	False AI accuracy claims
Air AI	Aug 2025	$18M judgment ($50K cash, rest suspended); ban (March 2026)	$19M	AI-powered earnings scheme + agentic AI claims

DoNotPay sold a “robot lawyer” service that the FTC alleged had never been tested against the work of human lawyers. The company paid $193,000 and accepted a ban on advertising legal services without substantiating the comparison. The case is the FTC’s clearest articulation that marketing an AI tool as a substitute for a licensed professional requires evidence the AI can actually do the work.

Ascend Ecom, Ecommerce Empire Builders, and FBA Machine were structurally similar: AI-branded business-opportunity schemes that promised passive income through e-commerce stores. Each was banned permanently from the industry. FBA Machine alone produced more than $15 million in consumer losses.

Click Profit, filed six months after the original sweep, recycled the same playbook at larger scale. The defendants charged customers between $45,000 and $75,000 to set up Amazon storefronts allegedly powered by AI and “affiliated with” Nike, Disney, and similar brands. The AI did not exist; the brand partnerships did not exist. Consumer losses exceeded $14 million. The August 2025 stipulated final order entered a $13.6 million monetary judgment and banned the operators permanently. The order also bars the defendants from any future business that “uses any automated, algorithmic, artificial intelligence, or machine learning process” — a notable expansion of remedy scope.

Workado is the most technically interesting recent case. The company sold an “AI Content Detector” tool that it advertised as 98% accurate at distinguishing AI-generated text from human-written text. Independent FTC testing found the actual accuracy rate was 53% — essentially a coin flip. The April 2025 proposed order (final August 21, 2025; Docket C-4822) required Workado to stop the accuracy claims, send corrective notice to consumers, and submit to compliance monitoring. There was no monetary penalty. The vote was 3-0.

Air AI was the most aggressive case in the cluster. The company sold what it described as agentic AI — autonomous customer-service representatives that could replace human staff — to small businesses for tens of thousands of dollars. The FTC alleged the AI did not perform as marketed, and that owners Caleb Maddix, Ryan O’Donnell, and Thomas Lancer made false earnings claims to extract additional payments. The August 2025 complaint sought a permanent ban. The March 24, 2026 stipulated permanent injunction entered an $18 million monetary judgment, suspended to $50,000 in cash relief based on inability to pay, and banned all three operators from marketing business opportunities, telemarketing, or making earnings claims without substantiation.

Across the eight active cases, the pattern is consistent: false earnings claims wrapped in AI marketing, plus one accuracy-claim case (Workado) and one professional-substitution case (DoNotPay). Operation AI Comply is functionally a deception sweep with AI as the recurring substrate.

Why was Rytr’s order set aside?

On December 22, 2025, the FTC voted 2-0 to set aside the December 2024 Rytr consent order. The set-aside was the first significant unwind of an Operation AI Comply case and the most direct application of the Trump administration’s July 2025 AI Action Plan, which directs the FTC to “review all FTC final orders, consent decrees, and injunctions, and, where appropriate, seek to modify or set-aside any that unduly burden AI innovation.”

The original Rytr order targeted an AI writing assistant that included a “Testimonial & Review” tool. Subscribers could enter limited input and receive detailed customer reviews back. The FTC alleged the tool produced reviews containing material details with no relation to the input — almost certainly false for any user who copied them and posted them online. The Commission found Rytr violated Section 5 in two ways: (1) by providing customers the “means and instrumentalities” to deceive consumers, and (2) by offering an unfair practice that risked polluting the marketplace with fake reviews. Then-Commissioners Ferguson and Holyoak dissented in 2024, arguing the theory swept too broadly.

The 2025 set-aside order, written from Ferguson’s now-majority position, is six pages of doctrinal recalibration. It argues that the original complaint failed to allege facts fitting the established applications of “means and instrumentalities” liability, and that “treating as categorically illegal a generative AI tool merely because of the possibility that someone might use it for fraud is inconsistent with our precedents and common sense.”

Three things are notable about the set-aside:

The legal theory, not the underlying conduct, is the issue. Rytr is not vindicated. If a Rytr subscriber actually deployed thousands of fake reviews, the FTC could still pursue that subscriber for unfair or deceptive practices. The set-aside narrows liability for tool-makers, not end users.
Rytr consented to vacating the order. The set-aside was bilateral. Rytr waived rights under Rule 3.72(b) of the Commission’s Rules of Practice. The Commission did not have to litigate the underlying merits.
The vote was 2-0. Both Republican Commissioners voted to set aside. The Democratic minority did not participate (Commissioner Slaughter had departed; Commissioner Bedoya did not vote). The set-aside reflects the current composition of the agency, which may shift as appointments change.

For practitioners, the operative takeaway is narrow: marketing a general-purpose AI tool that bad actors might misuse does not, by itself, trigger Section 5 liability. Marketing an AI tool by misrepresenting what it does — the DoNotPay, Click Profit, Workado, Air AI pattern — still does.

What is algorithmic disgorgement?

Algorithmic disgorgement is the FTC’s most powerful AI-specific remedy. The agency has ordered defendants to delete not just improperly collected data, but the AI models and algorithms trained on that data. It has done this four times: Everalbum (January 2021), WW International / Kurbo (March 2022), Rite Aid (December 2023), and Amazon Alexa (May 2023).

Case	Year	Data Deleted	Models/Algorithms Deleted
Everalbum / Paravision	2021	Consumer photos collected without consent	All facial recognition models trained on those photos
WW International (Kurbo)	2022	Children’s data collected in violation of COPPA	All AI/ML models trained on children’s data
Rite Aid	2023	Eight years (2012-2020) of facial recognition data	All algorithms used in the shoplifting-detection program
Amazon Alexa	2023	Children’s voice recordings retained indefinitely	Models trained on children’s voice data

The remedy is consequential because AI models are increasingly the most valuable asset a company holds. A trained model represents months or years of data collection, compute, and refinement. A disgorgement order can erase that investment in a single afternoon.

Algorithmic disgorgement remains available under the current administration. It was not addressed in the Rytr set-aside or the March 2026 AI Policy Statement, both of which focused on deception and unfairness liability rather than remedies. Companies should assume the remedy will be applied to any AI system trained on improperly collected data.

The Rite Aid case is the most cited because it is the FTC’s first explicit algorithmic discrimination action. The chain deployed facial recognition in retail pharmacies between 2012 and 2020 without testing for accuracy or demographic disparity. The system generated thousands of false-positive matches concentrated among women and people of color. The 5-year facial recognition ban plus full data and model deletion remains the agency’s most aggressive AI remedy on the public record.

What patterns emerge from the cases?

Six patterns recur across the 8 Operation AI Comply cases plus the four algorithmic disgorgement actions:

Earnings claims plus AI marketing is the highest-risk combination. Five of the 8 Operation AI Comply cases (Ascend Ecom, EEB, FBA Machine, Click Profit, Air AI) involved AI-branded business opportunities that promised income the operators could not deliver. The Business Opportunity Rule and the Consumer Review Fairness Act provide additional theories beyond Section 5.
Accuracy claims must be substantiated. Workado promised 98% accuracy and could substantiate 53%. DoNotPay promised lawyer-equivalent service and could substantiate nothing. The pattern is consistent: if you advertise a measurable AI capability, the FTC expects you to have measurement evidence on file.
Substitution claims are scrutinized intensely. Marketing an AI tool as a substitute for a licensed professional (DoNotPay) or a trained employee (Air AI’s “autonomous sales staff”) triggers a higher evidence standard. The closer your marketing language gets to “replaces a human,” the more substantiation you need.
Algorithmic discrimination is unfairness. Rite Aid established that deploying biased AI without reasonable safeguards constitutes “unfairness” under Section 5. The Meta housing-ads case (DOJ, 2022) extended this to ad delivery algorithms. The Workday class certification (N.D. Cal. 2025) extended it to AI vendors as employer agents. Companies that deploy AI in employment, housing, lending, or healthcare without disparate-impact testing face escalating exposure.
Means and instrumentalities is now narrow. After the Rytr set-aside, the FTC will not pursue tool-makers solely on the theory that their tools could be misused. Liability requires misrepresentation by the tool-maker about what the tool itself does.
Bipartisan continuity on demonstrable harm. Both the Khan-era FTC and the Ferguson-era FTC have brought AI cases. The cases that succeeded under both administrations involved clear consumer injury tied to specific false claims. The cases that drew dissent or were set aside (Rytr, the 2024 Khan-era 6(b) study on AI partnerships) involved more attenuated theories.

What is the FTC likely to target next?

Three forward-looking signals shape the 2026-2027 enforcement horizon:

The March 11, 2026 AI Policy Statement introduces a graduated enforcement schedule. Through 2026, the FTC will issue consent orders and guidance letters in a “warning phase.” Starting 2027, full enforcement begins, with civil penalties up to $53,088 per violation under Section 19 of the FTC Act for violations of subsequent orders.

The March 9, 2026 Trump executive order creates an AI Litigation Task Force in the DOJ and directs the Commerce Department to survey state AI laws for federal preemption challenges. The Policy Statement signals the FTC’s view that some state-mandated AI bias corrections may themselves be deceptive under federal law. Companies operating across state lines should expect federal-state preemption litigation to escalate.

Agentic AI is the next probable enforcement frontier. Air AI was the first FTC case to explicitly address agentic AI claims (autonomous customer-service representatives). Singapore released the world’s first agentic AI guidance in January 2026. The FTC has signaled, both in its March 19-20 conference and in the AI Policy Statement, that AI agents acting on consumer accounts will receive heightened scrutiny. Marketing an “AI agent” that can take actions on behalf of users requires substantiation of (1) what actions the agent can actually take, (2) what error rates the agent has, and (3) what consumer recourse exists when the agent makes mistakes.

Expect 2026-2027 cases to focus on AI hiring tools (post-Workday class certification), AI in healthcare diagnosis (post-Pieces Technologies state AG settlement), and agentic AI products marketed to consumers and small businesses.

How should companies respond? Six steps to avoid FTC scrutiny

The same compliance moves work under both Khan-era and Ferguson-era enforcement, because both administrations agree on the core principle: false claims about AI capability are unlawful, and substantiation must exist before the claim is made.

Substantiate every measurable AI claim before publishing it. If you advertise accuracy, speed, or capability, document the testing methodology, the dataset, and the result. Workado’s downfall was claiming 98% when its own data showed 53%. Internal QA must align with external marketing.
Drop substitution language unless you can prove it. “AI lawyer,” “replaces your sales team,” and “fully autonomous” are claims with maximum legal exposure. If your AI is a tool that augments human work, market it that way. If you genuinely replace a licensed professional, you need licensure evidence, not just performance evidence.
Test for algorithmic disparate impact in regulated domains. Employment, housing, lending, healthcare, and education AI face active enforcement (Workday, Sirius XM, HireVue, Mobley v. Workday). Run disparate-impact testing under Title VII, FHA, and ECOA frameworks before deployment, not after a complaint.
Document data provenance and obtain meaningful consent. Algorithmic disgorgement is the FTC’s nuclear option. The four cases that triggered it (Everalbum, WW/Kurbo, Rite Aid, Amazon Alexa) all involved data collected without consent appropriate to the eventual use. NIST AI Risk Management Framework documentation and an internal data-provenance log are the minimum defenses.
Build a substantiation file for agentic AI claims. If you market AI agents that take actions, document (1) the action set the agent is authorized to perform, (2) the error rate per action class, (3) the human-review thresholds, and (4) the consumer recourse path when the agent errs. Air AI’s failure on each of these dimensions made the case unwinnable.
Treat the FTC AI Policy Statement as your enforcement calendar. 2026 is the warning phase. Use it to fix what you cannot defend. Companies that arrive at 2027 without substantiated claims, disparate-impact testing, and consent documentation face civil penalties up to $53,088 per violation under post-Policy-Statement consent orders.

For deeper context on adjacent enforcement, see our coverage of the SEC’s parallel AI washing actions, AI liability in the United States, and the NIST AI Risk Management Framework that the FTC increasingly cites in consent orders. State-level enforcement parallels the FTC pattern in Texas TRAIGA and Colorado AI Act 2026. For comparative perspective, see EU vs US AI Regulation: The Definitive Comparison.

Sources

FTC. “FTC Announces Crackdown on Deceptive AI Claims and Schemes.” September 25, 2024. https://www.ftc.gov/news-events/news/press-releases/2024/09/ftc-announces-crackdown-deceptive-ai-claims-schemes
FTC. “FTC Reopens and Sets Aside Rytr Final Order in Response to the Trump Administration’s AI Action Plan.” December 22, 2025. https://www.ftc.gov/news-events/news/press-releases/2025/12/ftc-reopens-sets-aside-rytr-final-order-response-trump-administrations-ai-action-plan
FTC. “Order Reopening and Setting Aside Order — Rytr LLC.” December 22, 2025. https://www.ftc.gov/system/files/ftc_gov/pdf/Rytr-Order.pdf
FTC. “FTC Acts to Stop ‘Click Profit’ Online Business Opportunity that Has Cost Consumers At Least $14 Million.” March 18, 2025. https://www.ftc.gov/news-events/news/press-releases/2025/03/ftc-acts-stop-click-profit-online-business-opportunity-has-cost-consumers-least-14-million
FTC. “Stipulated Order for Permanent Injunction and Monetary Judgment — Click Profit.” August 26, 2025. https://www.ftc.gov/system/files/ftc_gov/pdf/ClickProfit-StipulatedFinalOrder-Emslie-et-al.pdf
FTC. “FTC Order Requires Workado to Back Up Artificial Intelligence Detection Claims.” April 28, 2025. https://www.ftc.gov/news-events/news/press-releases/2025/04/ftc-order-requires-workado-back-artificial-intelligence-detection-claims
FTC. “Air AI and its Owners will be Banned from Marketing Business Opportunities.” March 24, 2026. https://www.ftc.gov/news-events/news/press-releases/2026/03/air-ai-its-owners-will-be-banned-marketing-business-opportunities-settle-ftc-charges-company-misled
FTC. “FTC Finalizes Order with DoNotPay; Prohibits Deceptive AI Lawyer Claims, Imposes Monetary Relief, Requires Notice.” February 2025. https://www.ftc.gov/news-events/news/press-releases/2025/02/ftc-finalizes-order-donotpay-prohibits-deceptive-ai-lawyer-claims-imposes-monetary-relief-requires
FTC. “Rite Aid Banned from Using AI Facial Recognition.” December 19, 2023. https://www.ftc.gov/news-events/news/press-releases/2023/12/rite-aid-banned-using-ai-facial-recognition-after-ftc-says-retailer-deployed-technology-without
FTC. “FTC, DOJ Charge Amazon with Violating Children’s Privacy Law by Keeping Kids’ Alexa Voice Recordings Forever.” May 31, 2023. https://www.ftc.gov/news-events/news/press-releases/2023/05/ftc-doj-charge-amazon-violating-childrens-privacy-law-keeping-kids-alexa-voice-recordings-forever
FTC. “Everalbum, Inc., In the Matter of.” January 8, 2021. https://www.ftc.gov/legal-library/browse/cases-proceedings/192-3172-everalbum-inc-matter
Inside Privacy. “FTC Sets Aside Rytr Final Order Pursuant to White House AI Action Plan.” January 5, 2026. https://www.insideprivacy.com/united-states/federal-trade-commission/ftc-sets-aside-rytr-final-order-pursuant-to-white-house-ai-action-plan/
Reed Smith / Lexology. “Rewriting Rytr: The FTC sets aside a Final Order to implement America’s AI Action Plan.” December 23, 2025. https://www.lexology.com/library/detail.aspx?g=ce756459-956a-4582-a906-8adfc9bd0f7d
Mondaq / Benesch. “One Year In, FTC’s ‘Operation AI Comply’ Continues Under New Administration.” October 23, 2025. https://www.mondaq.com/unitedstates/new-technology/1695094/one-year-in-ftcs-operation-ai-comply-continues-under-new-administration-signaling-enduring-enforcement-focus
White House. “America’s AI Action Plan.” July 2025.
Executive Order: “Ensuring a National Policy Framework for Artificial Intelligence.” December 2025; March 9, 2026 amendment.

Reg Intel is not a law firm and does not provide legal services. This article is for informational purposes only and should not be relied upon as legal advice. Consult qualified counsel for your specific compliance situation.