Skip to content
Menu
Finabase Learn

Hedge funds and AI · AI in investing

AI-washing: SEC cases over false AI claims and what to ask

What AI-washing is, the US cases against firms that overstated their AI from 2024 to 2026, the rules behind them, and the questions to ask any product that claims AI.

On this page
  1. What counts as AI-washing
  2. The rules behind the cases
  3. The cases, in order
  4. What the cases have in common
  5. Questions to ask a product that claims AI
  6. Where the rules fall short
  7. How Finabase does it
  8. Frequently asked questions
  9. Summary
  10. Check your understanding

AI-washing is claiming to use artificial intelligence that you do not use, or claiming it does more than it does, to win customers or investors. The term borrows from "greenwashing", the practice of overstating a product's environmental credentials. In investing it shows up in three places: advisers and apps that say their picks come from AI, companies that tell shareholders their product runs on AI, and outright frauds that use "AI" as the story behind a scheme.

US regulators have treated these claims as an ordinary question of truth in marketing and disclosure. There is no special AI law behind the cases; the rules that forbid any false statement to clients or investors apply to statements about AI too. This lesson walks through the cases brought from March 2024 to September 2026, the rules they rest on, what they have in common, and the questions an investor can ask any product that says it uses AI.

What counts as AI-washing

The cases fall into a few patterns:

  • AI that does not exist. The firm says a model drives its decisions, and none does.
  • AI that does less than claimed. A model exists, but people do much of the work, or it covers only part of the process the marketing describes.
  • AI as a credential. "The first regulated AI advisor" or "AI-driven forecasts" used as a selling point without anything to back it up.
  • AI as the cover story for a fraud. A trading bot or signal service that does not trade at all. The scam side of this is covered in AI trading bot and signal group scams: the pitches and the tells.

Part of the difficulty is that "AI" has no fixed meaning. The Financial Stability Oversight Council noted in its 2023 annual report that definitions vary and that there is not always a sharp line between AI and older statistical modeling; some regression techniques in use for decades could arguably be called AI. That vagueness is what makes the word useful to a marketer, and why the question to ask is always what the system actually does.

The rules behind the cases

Which rule applies depends on who made the claim.

Investment advisers. A registered adviser is a fiduciary, and section 206 of the Investment Advisers Act forbids fraud on clients. Section 206(2) does not require intent; negligence is enough, which matters when a firm simply failed to check its own claims. Under section 206(4), the SEC's Marketing Rule, 17 CFR 275.206(4)-1, sets out what an adviser's advertisements may not do. Among other things, an advertisement may not:

  1. include an untrue statement of a material fact, or leave out a fact needed to make a statement not misleading;
  2. include a material statement of fact that the adviser has no reasonable basis to believe it can substantiate on demand by the SEC;
  3. suggest an untrue or misleading inference about the adviser;
  4. talk up potential benefits without fair and balanced treatment of the material risks and limitations;
  5. present performance results, or the periods they cover, in a way that is not fair and balanced.

The substantiation test is the one that catches AI claims. If an adviser's website says its forecasts are "AI-driven", it must be able to show the SEC that they are. The current version of the rule was adopted in 2020, with compliance required from November 4, 2022.

Public companies. A company's statements in its filings and to the market are covered by the Securities Act and the Exchange Act. Section 17(a)(2) of the Securities Act reaches misleading statements made in offering securities, and section 13(a) of the Exchange Act requires accurate periodic reports.

Anyone raising money. Rule 10b-5 under the Exchange Act forbids fraud in connection with buying or selling any security, which covers start-ups raising venture money as well as listed companies. When the facts are bad enough, the Justice Department can also bring criminal charges, such as securities fraud and wire fraud.

The cases, in order

US cases over false AI claims, March 2024 to September 2026 Date Who What was claimed Outcome Mar 2024 Delphia (USA) Used AI and machine learning on client data Settled; $225,000 penalty Mar 2024 Global Predictions "First regulated AI financial advisor" Settled; $175,000 penalty Oct 2024 Rimar Capital AI platform trading client accounts Settled; $310,000 in penalties Jan 2025 Presto Automation Drive-thru AI that needed no people Settled; no penalty Apr 2025 Nate, Inc. founder Shopping app that ran on AI SEC suit and criminal charges Aug 2026 GenesisAI Corp. Revenue and demand for an AI marketplace Settled; CEO $50,000 penalty Sep 2026 Cryptoaiml, TSAI Pro AI trading signals; AI bots to rent SEC suits, pending
Selected US cases over false AI claims, March 2024 to September 2026. Sources: SEC press releases, orders and litigation releases, and the US Attorney for the Southern District of New York, as listed under Sources. Settled cases were resolved without admitting or denying the findings.

Delphia and Global Predictions (March 2024)

On March 18, 2024 the SEC settled charges against two registered advisers. Its chair at the time said: "Such AI washing hurts investors."

Delphia (USA) Inc., based in Toronto, said from about August 2019 to August 2023, in its Form ADV brochures, a press release and on its website, that it used artificial intelligence and machine learning on data its clients shared to make its investment decisions more accurate. The SEC found it never used client data as an input to its investing algorithms. During an SEC examination, Delphia acknowledged in July 2021 that it had not used client data and had not built an algorithm to do so, and it corrected its brochure in August 2021, yet similar claims continued in other advertising until August 2023. The order found violations of sections 206(2) and 206(4) and of the Marketing Rule and the compliance-program rule. Delphia was censured and paid a $225,000 penalty.

Global Predictions Inc., of San Francisco, ran an online advice service and called itself the "first regulated AI financial advisor", with "expert AI-driven forecasts". The SEC found those statements false and misleading and that the firm could not substantiate them. The same order found it had advertised tax-loss harvesting it did not offer, could not back up a claim that its models beat IMF forecasts by 34%, showed hypothetical performance to the general public without the required policies, and put a misleading liability disclaimer, or hedge clause, in its client contract. It paid $175,000.

Rimar Capital (October 2024)

On October 10, 2024 the SEC settled charges against Rimar Capital USA, Rimar Capital LLC, its owner Itai Liptz and board member Clifford Boro. They raised nearly $4 million from 45 investors, the SEC found, partly by describing an AI-driven platform that traded client accounts automatically, along with false figures for assets under management and returns. Liptz also used company money for personal expenses. The penalties totaled $310,000 ($250,000 for Liptz, $60,000 for Boro), and Liptz paid $213,611 in disgorgement and interest and was barred from the industry with a right to reapply after five years.

Presto Automation (January 2025)

The SEC settled with Presto Automation, a company that reports to the SEC and sells drive-thru ordering technology to restaurants, on January 14, 2025. From November 2021 to May 2023, the SEC found, Presto did not disclose that for a period the speech-recognition technology in its AI ordering product was owned and operated by a third party. When it moved to its own technology, it said the product removed the need for human order takers, when most orders still needed a person. The SEC imposed no penalty, citing Presto's cooperation and remedial steps.

Nate (April 2025)

On April 9, 2025 the US Attorney in Manhattan announced criminal charges against Albert Saniger, founder and former chief executive of Nate, Inc., and the SEC filed a parallel civil suit. The indictment alleges that he raised more than $40 million (the SEC puts it above $42 million) by telling investors that Nate's shopping app completed purchases with AI and without human intervention, when its automation rate was effectively zero and hundreds of contract workers in a call center in the Philippines completed the purchases by hand. He was charged with one count each of securities fraud and wire fraud, each carrying a maximum of 20 years in prison. These are allegations; he is presumed innocent unless proven guilty.

2026: an AI start-up and two alleged bot frauds

On August 26, 2026 the SEC filed a settled case against GenesisAI Corp., a former Miami AI start-up, and its founder. According to the SEC, they raised more than $5.3 million from over 4,000 investors through crowdfunding and Regulation A offerings by overstating projected revenue, valuation and demand for a marketplace for AI models that was never commercially viable, including a customer waitlist that did not exist. Subject to court approval, the founder agreed to pay $50,000 in disgorgement, $9,184.53 in interest and a $50,000 penalty, without admitting or denying the allegations.

On September 29, 2026 the SEC sued Cryptoaiml and TSAI Pro, groups it says are probably run from overseas. It alleges that one sent "AI-generated trading signals" in WhatsApp groups and steered investors to a fake platform, and the other rented out AI trading bots that did not exist, with falsified Form D filings and a fake SEC certificate to look legitimate. The SEC says more than $12.5 million and $2.8 million were misappropriated. Those cases were pending when this lesson was written.

What the cases have in common

  • The claim was checkable, and nobody checked. Each claim was a statement of fact about what a system did, and each turned out to be false when an examiner, a court or an employee looked.
  • People did the work. In Presto and Nate, humans did much of what the marketing credited to AI. In Delphia, the promised use of client data never happened.
  • AI was rarely the only problem. Global Predictions also misstated its services and performance; Rimar also misstated assets and returns. A firm loose with one claim tends to be loose with others.
  • The penalties were small. The three settled cases against advisers and their owners cost $710,000 in penalties in total. The deterrent is the public order, not the fine.

Questions to ask a product that claims AI

Questions to ask a product that says it uses AI Ask A real answer sounds like What exactly does the AI do? One task, named: "ranks news by relevance", not "powers everything" What data does it use, and how old? Named sources and dates; how current prices and filings get in Does a person check its output? Who reviews what, and what happens when it is wrong Show me results, losses included Real accounts or clearly labelled hypothetical results, all periods Where is this in your filings? The same claim in Form ADV, a prospectus or a 10-K Who is registered, and with whom? A firm and people you can find on BrokerCheck or the SEC search
Six questions for any investing product that says it uses AI, and the kind of answer that suggests the claim is real.
  1. What exactly does the AI do? "It ranks news by relevance" or "it drafts a summary a person reviews" is an answer. "AI powers our whole process" is not.
  2. What data does it use, and how current is it? Ask where prices and filings come from and how often they update. A model trained once and never refreshed is describing the past.
  3. Who checks its output? A real system has people who review what it produces and a process for when it is wrong.
  4. Show me results, losses included. Ask for actual results over all periods, not a chosen window or a hypothetical backtest. Overfitting and backtest bias: 4,094 rules tested on SPY explains why a perfect backtest is a warning, not a credential.
  5. Where is this in your filings? A registered adviser describes its methods in its Form ADV brochure; a fund in its prospectus; a public company in its 10-K. If the AI is central to the pitch but missing from the filings, ask why.
  6. Who is registered, and with whom? Look the firm and its people up on BrokerCheck and the SEC's adviser search (How to check a broker or adviser: BrokerCheck, IAPD and Form CRS). Registration does not prove a claim is true, as Delphia and Global Predictions show, but no registration at all is a much bigger warning.

Where the rules fall short

  • Enforcement comes after the harm. Delphia's claims ran for about four years; the cases are warnings for the next investor more than protection for the last one.
  • Settlements are not admissions. Every settled case here was resolved without admitting or denying the findings, which is standard SEC practice.
  • Private and offshore operators are hard to reach. Start-ups raising private money and groups run from abroad, such as those in the September 2026 suits, may have little left to recover.
  • Priorities move. In February 2025 the SEC created a Cyber and Emerging Technologies Unit of about 30 specialists, listing fraud using AI among its priorities, and its examiners said for fiscal 2026 that they would review advisers' and brokers' claims about their AI for accuracy. The number of cases still varies from year to year, so a product's claims are yours to check whatever the enforcement climate.

How Finabase does it

Finabase's AI features are described by what they do, and the guides say what they cannot do. On the web, Ask Finabase AI: research stocks with the AI chat on the web answers questions with Finabase's market data shown in data cards beside the answer, has a daily request limit, and states that AI can make mistakes and that nothing it says is investment advice. In Stock Master, the AI tab writes a short analysis of a stock, labelled as AI-generated, for information only and possibly mistaken (Trend Score and AI analysis in Stock Master). Finabase does not manage money, run trading bots or sell signals. Using AI chatbots for stock research: how to check what they say explains how to check any chatbot's answer, including ours.

Frequently asked questions

What is AI-washing?

Claiming to use AI that you do not use, or overstating what it does, to attract clients or investors. It is the AI version of greenwashing, and in the US it is pursued under ordinary antifraud and disclosure rules.

Is AI-washing illegal?

A false or unsupported statement of fact to clients or investors can break the securities laws whatever its subject. For registered advisers, the Marketing Rule also requires that factual claims in advertising can be substantiated when the SEC asks.

What was the first SEC AI-washing case?

The SEC's settled actions against Delphia (USA) Inc. and Global Predictions Inc., announced on March 18, 2024, with $400,000 in penalties between them.

Does a registered firm's AI claim mean it is true?

No. Both March 2024 cases involved registered advisers. Registration means a regulator can examine and sanction the firm; it does not mean someone has checked its marketing.

Summary

AI-washing is an old problem, a false claim, in a new vocabulary. US regulators have handled it with existing rules: the Marketing Rule's demand that advisers be able to prove what they advertise, antifraud provisions for companies and fund-raisers, and criminal charges where the deception was deliberate. The cases from 2024 to 2026 share a pattern of checkable claims that nobody checked, often with people doing the work credited to AI. Before trusting any product that says it uses AI, ask what it does, on what data, who checks it, where the claim appears in its filings and what its full record looks like.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. An adviser's website says its portfolios are built by "proprietary AI forecasts". Under the SEC's Marketing Rule, what must be true?

    1. The adviser must have a reasonable basis to believe it can substantiate the claim if the SEC asks
    2. Nothing, because "AI" has no legal definition
    3. The adviser must register its AI model with the SEC before advertising it
    Show the answer

    A. The adviser must have a reasonable basis to believe it can substantiate the claim if the SEC asks The Marketing Rule bars a material statement of fact that the adviser cannot reasonably expect to substantiate on demand. "Our forecasts are AI-driven" is a factual claim. Global Predictions was found unable to back up exactly this kind of claim. There is no separate registration of models.

  2. Delphia told clients it used AI and machine learning on their data. What did the SEC find?

    1. It never used client data as an input to its investing algorithms, and kept making the claim in some advertising after an exam
    2. Its AI worked as described, but it charged too much for it
    3. It used the data, but without asking clients for permission
    Show the answer

    A. It never used client data as an input to its investing algorithms, and kept making the claim in some advertising after an exam The order found Delphia had not built or used the client-data models it advertised. It admitted this to examiners in July 2021 and corrected its brochure, but similar claims continued in other advertising until August 2023. It paid a $225,000 penalty.

  3. In the Presto Automation and Nate cases, what was the gap between the claim and the reality?

    1. People did much of the work that the companies credited to AI
    2. The AI worked, but the companies overstated their revenue from it
    3. The companies used AI without disclosing it to customers
    Show the answer

    A. People did much of the work that the companies credited to AI Presto said its drive-thru product removed the need for human order takers when most orders still needed a person. Nate is alleged to have had an automation rate of effectively zero, with contract workers completing purchases by hand. Asking who checks or does the work is one of the best tests of an AI claim.

  4. A firm is registered with the SEC and its marketing leans heavily on AI. What does the registration tell you about the AI claims?

    1. Only that a regulator can examine and sanction the firm; the claims still need checking
    2. That the SEC has reviewed and approved its marketing
    3. That its AI has been tested by the SEC
    Show the answer

    A. Only that a regulator can examine and sanction the firm; the claims still need checking Both of the first SEC AI-washing cases, Delphia and Global Predictions, involved registered advisers. Registration makes a firm accountable; it is not an endorsement of anything it advertises. Look for the claim in its Form ADV and ask for a full record.

Sources and further reading

  1. SEC — Press release 2024-36: SEC charges two investment advisers with making false and misleading statements about their use of artificial intelligence (March 18, 2024)
  2. SEC — In the Matter of Delphia (USA) Inc., Advisers Act Release IA-6573 (March 18, 2024)
  3. SEC — In the Matter of Global Predictions, Inc., Advisers Act Release IA-6574 (March 18, 2024)
  4. SEC — Press release 2024-167: settled charges against Rimar Capital entities and owner for defrauding investors in an "AI washing" scheme (October 10, 2024)
  5. SEC — Administrative proceeding summary, Presto Automation Inc., Release 33-11352 (January 14, 2025)
  6. US Attorney, Southern District of New York — Press release 25-082: Tech CEO charged in artificial intelligence investment fraud scheme (April 9, 2025)
  7. SEC — Litigation Release 26282, SEC v. Albert Saniger (April 11, 2025)
  8. SEC — Litigation Release 26619, SEC v. GenesisAI Corp. and Archil Cheishvili (August 26, 2026)
  9. SEC — Litigation Release 26654, SEC v. Cryptoaiml Ltd. et al. and SEC v. TSAI Pro Ltd. et al. (September 29, 2026)
  10. eCFR text via Cornell LII — 17 CFR 275.206(4)-1, Investment adviser marketing (the Marketing Rule)
  11. SEC — Press release 2025-42: SEC announces Cyber and Emerging Technologies Unit (February 20, 2025)
  12. SEC Division of Examinations — Fiscal Year 2026 Examination Priorities
  13. Financial Stability Oversight Council — 2023 Annual Report, section 3.3.3 (definitions of AI)