The SEC’s AI-washing cases against two advisers punished one plain thing: describing AI capabilities the firms did not have.
What AI-washing means
AI-washing is the practice of describing a firm’s use of artificial intelligence in a way that overstates what the firm actually does. The term borrows from greenwashing. In both cases the problem is the gap between the statement and the fact, not the underlying activity.
The SEC gave the term its enforcement meaning in settled cases against two advisers, Delphia (USA) Inc. and Global Predictions Inc., over false and misleading statements about how they used artificial intelligence. The firms agreed to settle the charges and to pay $400,000 in total civil penalties to the SEC. SEC Chair Gary Gensler said at the time: “Investment advisers should not mislead the public by saying they are using an AI model when they are not.” He called the conduct AI washing and said it hurts investors.
For a compliance officer the point is narrow and useful. The cases did not punish using AI. They did not punish using AI badly. They punished statements. Every theory of liability in the two orders runs through a sentence the firm published somewhere: a filing, a press release, a website, a social media post, an advisory contract. That is the lens to carry through the rest of this guide.
What the SEC found at Delphia
According to the SEC’s order, from 2019 to 2023 the Toronto-based firm made false and misleading statements in its SEC filings, in a press release and on its website about its purported use of AI and machine learning that incorporated client data in its investment process. The order cites Delphia’s claim that its artificial intelligence could “predict which companies and trends are about to make it big and invest in them before everyone else.” The order’s finding is that the statements were false and misleading: the firm did not have the AI and machine learning capabilities it described.
Two features of this finding matter for other firms. First, the statements appeared in several channels, including SEC filings. A marketing claim that migrates into a Form ADV brochure carries the same obligation of accuracy and a wider audience. Second, the claim was specific and testable: it described a capability, a data input and a predicted outcome. A specific claim is easy to compare against what the firm actually runs, and in Delphia’s case the comparison went against the firm.
Delphia was also charged under the Marketing Rule. Among other things, that rule bars a registered adviser from putting out an advertisement containing an untrue statement of material fact. In our reading, this is the charge most likely to recur in future AI cases against advisers, because it does not require the SEC to prove anything about the technology. It requires only an advertisement and an untrue statement.
What the SEC found at Global Predictions
The Global Predictions order concerns statements the San Francisco firm published in 2023 on its website and on social media. The SEC found that those claims about the firm’s use of AI were false and misleading. The SEC found that the firm falsely claimed to be the first regulated AI financial advisor. The SEC also found that it misrepresented the AI-driven forecasts its platform provided and that it violated the Marketing Rule.
The Global Predictions order is a reminder that AI claims rarely travel alone. The Marketing Rule finding rested on a false claim that the firm offered tax-loss harvesting services, and the order also found an impermissible liability hedge clause in its advisory contract, among other securities law violations. Once staff are reading a firm’s website and contracts for an AI claim, they are reading everything else on the page too. A firm that reviews its AI language in isolation is likely to miss the adjacent problems that an examiner or investigator would not.
What the firms agreed to
Without admitting or denying the SEC’s findings, both firms consented to orders finding that they violated the Advisers Act, censuring them and ordering them to cease and desist from violating the charged provisions. Under the orders, Delphia paid $225,000 and Global Predictions paid $175,000.
Staff of the SEC’s Division of Examinations assisted with the investigations. That detail suggests the route from an examination finding to an enforcement referral is short for this kind of statement. A firm should assume that language on its website could be compared against its actual systems during a routine examination, not only during an investigation.
What the cases did not decide
Readers over-read enforcement headlines, so it is worth being precise about what the orders did not do. They did not create a rule about AI. They did not set standards for model governance, model validation or disclosure of AI use. They did not say that an adviser must describe its algorithms in any particular way, or at all. They did not penalize either firm for a model that underperformed.
What they did was apply existing antifraud and advertising provisions to a particular subject matter. Gurbir S. Grewal, then Director of the SEC’s Division of Enforcement, put it this way: “if you claim to use AI in your investment processes, you need to ensure that your representations are not false or misleading.” Grewal added that public issuers making claims about their AI adoption must also remain vigilant about similar misstatements that may be material to individuals’ investing decisions. Gensler also said that when technologies come along they can create buzz from investors as well as false claims by those purporting to use them. Alongside the cases, the SEC’s Office of Investor Education and Advocacy issued an Investor Alert on artificial intelligence for the investing public.
Taken together, the statements signal that the SEC treats AI claims the way it has treated other fashionable claims: as a place where the gap between marketing and reality tends to open, and where the existing rules already reach. The obligations that applied before these cases still apply. An adviser’s fiduciary duty, the Advisers Act antifraud provisions and the Marketing Rule all govern AI statements exactly as they govern any other statement about the firm’s services.
The cases did not punish using AI; they punished statements about AI that the firms could not back up.
Where AI statements live
The two orders between them reached SEC filings, a press release, websites, social media and an advisory contract. A review that covers only the marketing deck is incomplete. The table below sets out the channels and the question to ask of each.
| Channel | Question to ask |
|---|---|
| Form ADV and other SEC filings | Does every sentence about AI or models describe something the firm runs today? |
| Website and landing pages | Could the firm demonstrate each AI claim to an examiner from its own systems? |
| Social media and press releases | Has anyone compared the superlatives against the product as built? |
| Pitch decks and RFP responses | Are vendor capabilities described as the firm’s own? |
| Client agreements | Do the services described match the services delivered, and are the terms permissible? |
| Investor letters and performance commentary | Is a human process described as automated, or vice versa? |
A checklist for describing AI use
Use this before any statement about AI leaves the firm.
- The statement names a capability the firm can demonstrate from its own systems today, not a roadmap item.
- Someone in technology or investment has confirmed in writing that the description is accurate.
- Vendor tools are described as vendor tools, and the firm does not claim ownership of what it licenses.
- Words such as “predict”, “proprietary”, “first” and “AI-driven” each have a specific fact behind them, or they are removed.
- The description of data inputs, including any client data, matches what the model actually ingests.
- The statement is consistent across filings, website, social media and contracts.
- The statement has been reviewed under the Marketing Rule where it is an advertisement.
- A dated record of the review is kept, so the firm can show what it believed and why when the statement was made.
Implementation steps
The checklist is the control. The steps below are how to put it in place.
- Inventory every statement about AI, machine learning, algorithms or automation across all channels in the table above. Include archived web pages and old press releases that are still reachable.
- For each statement, identify the person who can verify it and ask for a plain-language description of what the system does. Compare the two. Where the statement is broader than the system, rewrite or remove it.
- Add AI claims to the firm’s advertising review procedure. The Marketing Rule’s prohibition on untrue statements of material fact applies to them in the same way as to performance or testimonials.
- Put a change control step between product and marketing. When a model is retired, scaled back or replaced by a vendor, marketing should hear about it before the website goes out of date.
- Train the people who write for the firm. The problem in both orders began with a sentence someone drafted, likely in good faith, that outran the facts.
- Where a statement’s accuracy depends on a legal judgment, for example whether a contract clause is permissible, take legal advice. Taft does not provide it.
Questions firms ask
We use an off-the-shelf AI tool from a vendor. Can we say we use AI? A firm can describe what it does, accurately. Saying the firm uses a vendor’s tool for a stated purpose is a description. Saying the firm has built proprietary AI that predicts markets, when it has not, is the kind of statement the Delphia order found false and misleading. The safer sentence is the specific one.
Our AI claims are aspirational. Is that a defense? The orders turned on whether the capabilities existed when the statements were made. A statement framed in the present tense about something the firm intends to build is, in our reading, exactly the gap these cases describe. Write about plans as plans.
Does this apply to funds and issuers, or only to registered advisers? Both orders were against investment advisers and relied on the Advisers Act. Grewal’s statement extended the warning to public issuers describing their AI adoption. In our view the logic applies to any firm whose statements reach investors, whatever its registration.
Will examiners ask about this? Examination staff assisted with these investigations. That suggests examiners may compare a firm’s published AI claims with its systems, and we expect that line of questioning to be routine for firms that make such claims prominently. The source does not describe a dedicated examination initiative, so treat that as our expectation rather than a stated program.
What is the realistic exposure? The two firms together paid $400,000 in total civil penalties and accepted censures and cease-and-desist orders. The reputational cost of a public order describing a firm’s flagship claim as false is harder to measure and, in our view, larger.
Where Taft helps
Taft reviews the statements a firm makes about its use of AI against what the firm actually runs, and helps compliance teams build the review and change-control steps described above into their advertising procedures. Through Expert Line, compliance officers can put a specific question about an AI claim, a filing passage or a marketing review to someone who has held the compliance chair and get a direct answer. The full SEC announcement is available in the SEC press release. Taft is not a law firm and does not provide legal advice; where a statement’s accuracy depends on a legal judgment, the firm should take counsel’s view.
Questions
What is AI-washing?
AI-washing is describing a firm's use of artificial intelligence in a way that overstates what the firm actually does. The SEC used the term when it announced settled charges against Delphia (USA) Inc. and Global Predictions Inc. for false and misleading statements about their purported use of AI.
Did the SEC penalize the firms for using AI?
No. The orders found that the firms made false and misleading statements about AI capabilities they did not have. Delphia's order found it did not have the AI and machine learning capabilities it claimed, and Global Predictions was found to have falsely claimed to be the first regulated AI financial advisor.
How much did the firms pay?
The two firms agreed to pay $400,000 in total civil penalties, with Delphia paying a civil penalty of $225,000 and Global Predictions paying a civil penalty of $175,000 under the settled orders.
Which rules were charged?
Both firms consented to orders finding they violated the Advisers Act. Delphia was also charged under the Marketing Rule, which prohibits a registered investment adviser from disseminating an advertisement containing an untrue statement of material fact, and Global Predictions was found to have violated the Marketing Rule as well.
Sources
- SEC Charges Two Investment Advisers with Making False and Misleading Statements About Their Use of Artificial Intelligence, U.S. Securities and Exchange Commission
Taft does not provide legal advice. Content is for informational purposes only and subject to regulatory guidance.