AI washing: how the SEC and FTC police false AI claims

The short version

• AI washing means overstating what an AI system can do. It is a plain description of conduct, not a law of its own.
• US regulators reach it with existing law. The Securities and Exchange Commission (SEC) used the Investment Advisers Act and its Marketing Rule. The Federal Trade Commission (FTC) used its ordinary power over deception.
• On 18 March 2024 the SEC settled charges against two investment advisers, Delphia and Global Predictions, for false AI statements, with penalties of US$225,000 and US$175,000, each formally reprimanded (censured) and ordered to stop.
• In September 2024 the FTC’s Operation AI Comply brought five actions over deceptive AI claims, including the robot-lawyer matter against DoNotPay, which we cover below.
• The practical rule is the ordinary one. An AI claim has to be substantiated, and the regulator looks at the gap between the claim and what the firm could show, not at whether the technology is really AI.^1,2,3,6,7

What is AI washing?

AI washing means marketing a product or service as using artificial intelligence (AI), or as more capable than it is, when it does not have that capability. The term describes conduct; it is not a statute, and no law is titled AI washing. The conduct behind it is still reached under existing fraud and deception rules.¹

SEC officials used the label “AI washing” in the 18 March 2024 release. The term is formed by analogy to greenwashing, where a firm overstates how environmentally friendly it is; the release itself does not draw that comparison. The Director of the Division of Enforcement, Gurbir Grewal, framed the SEC’s work as protecting investors against AI washing, and the Chair, Gary Gensler, said advisers should not claim to use an AI model when they do not. Gensler put the harm plainly: such AI washing “hurts investors.” For a public company, the same release flagged the risk of a material misstatement, meaning a false statement about something important enough to affect an investor’s decision.¹

To substantiate a claim is to hold evidence that backs it up before the claim is made. That is the thread running through every matter below. The regulators did not ask whether the label AI was technically accurate. They asked whether the firm could back what it said. The general question of who is liable for what an AI system says is covered in our guide to AI in law.²

What has the SEC done about AI washing?

The clearest examples of AI washing come from the SEC, which has settled at least three enforcement matters over false AI statements, all under ordinary securities law. On 18 March 2024 it charged two investment advisers, Delphia and Global Predictions. On 10 October 2024 it charged a third adviser, Rimar, and two individuals.^1,4

From August 2019 to August 2023, Delphia (USA) Inc. told clients and regulators that it ran AI and machine learning over its retail clients’ spending and social-media data to shape its advice. The order finds it never fed that client data into its investment process at all. Delphia willfully violated Advisers Act sections 206(2) and 206(4), Rule 206(4)-1 (the Marketing Rule, which bars untrue statements in an adviser’s advertising) and Rule 206(4)-7 (the compliance rule). It was censured, ordered to cease and desist (to stop and not repeat the conduct), and paid a US$225,000 penalty.²

Global Predictions, Inc. claimed on its public website that its technology incorporated “expert AI-driven forecasts”, and claimed to be the first regulated AI financial advisor, a claim it could not back with documents when the SEC asked. It willfully violated Advisers Act sections 206(2) and 206(4), Rules 206(4)-1(a), (b) and (d), and Rule 206(4)-7. It was censured, ordered to cease and desist, and paid a US$175,000 penalty.³

The Rimar order reached two companies and two people. Rimar Capital, LLC purported to use artificial intelligence to perform automated trading for client accounts, and Rimar Capital USA, Inc. raised US$3.725 million from 45 investors through SAFEs (simple agreements for future equity, a contract that promises an investor equity later if certain events occur) on false statements about AI, assets under management and performance. The violations spanned Securities Act section 17(a), Exchange Act section 10(b) and Rule 10b-5, and Advisers Act sections 206(1) and 206(2). Itai Royi Liptz was ordered to pay a US$250,000 penalty and to give back US$202,604 in gains plus US$11,007.25 in interest (disgorgement and prejudgment interest), the disgorgement deemed satisfied by an offset against money he had earlier advanced to the firms. He was barred from the industry but may reapply after five years. Clifford Todd Boro was ordered to pay a US$60,000 penalty.^4,5

Each SEC matter settled without the firm admitting or denying the findings, so the orders are final but are not judicial precedent. The SEC did not rule on whether the technology deserved the AI label. It charged the distance between the claim and the substantiation.^2,4

How is the FTC acting against deceptive AI claims?

The FTC treats an overstated AI claim as ordinary deception under the FTC Act, and says the laws already in force apply to AI with no carve-out. Under the FTC’s long-standing deception standard, a claim is deceptive when it is likely to mislead a reasonable consumer about something material to their decision. In September 2024 the FTC announced Operation AI Comply, five actions against firms that used AI hype to deceive or that sold AI technology capable of being used to deceive.⁶

The Chair, Lina Khan, said “using AI tools to trick, mislead, or defraud people is illegal.” The five respondents were DoNotPay, Ascend Ecom, Ecommerce Empire Builders, Rytr and FBA Machine. The robot-lawyer matter against DoNotPay is covered in our guide to AI legal-advice chatbots and unauthorized practice.⁶

The clearest furnishing-the-means theory was Rytr, an AI writing assistant whose review feature could spin out long, frequently made-up customer reviews from a few words of input; the FTC said it furnished the means to produce fake reviews, on a split 3 to 2 vote. The proposed order bars Rytr from offering any service to generate consumer reviews or testimonials. Three of the five actions were business-opportunity schemes; two carried a headline dollar figure: Ascend Ecom, which the FTC said defrauded consumers of at least US$25 million, and FBA Machine, which it said cost consumers more than US$15.9 million.⁶

The FTC had signalled this before it sued. In February 2023 it published a business-guidance post, “Keep your AI claims in check”, by Michael Atleson of its Division of Advertising Practices. The official page returned an error to us on 11 November 2026, so we could not confirm its exact content against the FTC’s own text. The general point, that an AI claim must be backed by the same evidence as any other objective product claim, comes from secondary summaries and should be checked against the original post before anyone relies on it.⁷

What law do regulators use?

Regulators reach AI washing with the same rules they use for any false claim, and the common thread is substantiation. For the SEC, negligence is enough: a violation of Advisers Act section 206(2) needs no intent to deceive. For the FTC, Section 5 of the FTC Act covers unfair or deceptive acts or practices, and an AI claim is treated like any other.^2,6

State enforcers and statutes are moving too. Our guide covers the Colorado and Texas AI laws, and the EU AI Act. These are regulator actions; a competitor or a customer may have separate claims under advertising or contract law, which is outside this guide.

If you market AI, what makes a claim safe?

The rule is the ordinary advertising rule. Every AI claim must be substantiated before it is made, and it must match what the product actually does. The orders turn into a short checklist, grounded only in what they found.^2,3

• Say what the tool does now, not what you hope it will do. Delphia never fed client data into AI, yet it kept advertising that capability for years.
• Hold documents that back a superiority or first-or-only claim before you make it. Global Predictions could not produce documents to substantiate its first regulated AI financial advisor claim.
• Correct a claim fully once it is flagged, across filings, the website and social media. Delphia corrected some statements after an SEC exam but kept making others through August 2023, which the SEC counted as part of the violation.
• Keep written compliance policies that cover advertising on social media. Delphia had some advertising policies but lacked ones covering social media and a clear review-and-approval process, the gap Rule 206(4)-7, the compliance rule, is meant to close.
• Remember that a claim can mislead through an implied message or a demonstration, not only an openly false sentence. The FTC’s guidance makes the same point about performance and comparison claims.^2,3,7

A formal way to manage these claims is to run AI under a recognised framework. Our guide covers AI governance standards and what each one asks.

If you buy AI, how do you avoid buying AI washing?

To avoid buying AI washing, test the specific capability a vendor claims, the way the SEC tested Delphia’s and Global Predictions’ claims and found them false. Each item below is a diligence question drawn from an order.^2,3

• Ask the vendor to demonstrate the specific capability on your own data, not run a scripted demo.
• Ask what the model is actually trained on, and whether your data is used. Delphia’s claim to use client data was the core falsehood.
• Ask for substantiation of first, only, accuracy and performance claims. Global Predictions could not produce it when the SEC asked.
• Get the capability and its limits written into the contract’s representations, so a gap between the claim and the product is a breach you can point to.^2,3

A false claim about a product is a different problem from a false statement about a person. Our guide to AI defamation covers who answers when AI output invents something damaging about someone.

What does this mean for anyone building or buying AI?

In our view, what drives the enforcement so far is accountability: whether a firm could stand behind what it claimed. Nothing in these orders turns on whether a system is really AI. Each turns on the distance between what a firm claimed and what it could show. The SEC did not need a new statute. It used the Advisers Act and its Marketing Rule, and for a section 206(2) violation negligence was enough, with no intent to deceive required. The FTC did not need one either, treating an overstated AI claim as the deception it has always policed. Read together, the safest position is that the ordinary duty already governs AI marketing. A claim about what AI does has to be substantiated before it is made, and corrected fully once it is wrong. The rules and standards arriving now mostly make that duty explicit. Whoever makes the claim, or relies on it, stays answerable for it.^2,3,6

What is still open?

As of 11 November 2026, several threads are unresolved.

• The FTC policy statement that Executive Order 14365 called for, on accuracy in AI systems rather than on marketing claims, exists only as a proposed statement, published in the Federal Register on 7 July 2026. It is not a final rule. The order and its context are in our guide to AI in law.
• Operation AI Comply’s three business-opportunity matters, Ascend Ecom, Ecommerce Empire Builders and FBA Machine, were ongoing federal litigation when the sweep was announced. In each, the release says, a federal court issued an order temporarily halting the scheme and placing it under the control of a receiver (an outsider put in charge of a business’s assets). Their final judgments are not in the release we read.
• The settled SEC and FTC orders bind only their respondents and are not judicial precedent. The three SEC orders were entered without the firms admitting or denying the findings; the FTC settlements with DoNotPay and Rytr were proposed consent orders open to public comment when the sweep was announced.^6,8