How to Keep AI Claims Accurate Under the SEC Marketing Rule

Current as of October 29, 2026. This guide describes the SEC Marketing Rule, compliance rule and recordkeeping rule for investment advisers, two settled SEC orders, the SEC's 2025 withdrawal of its predictive data analytics proposal, SEC examination staff publications and an SEC web page and, for comparison, FINRA's communications rule and FAQ and a NASAA section publication, as published on that date. It is not legal advice; what your firm says in its marketing is a decision for your compliance team and counsel.

Advisers now describe AI in their marketing, and some descriptions run ahead of what the technology does. In March 2024 the SEC settled charges against two advisers, Delphia (USA) Inc. and Global Predictions Inc., over statements about their use of AI. Neither case needed a rule written for AI. The SEC Marketing Rule already says an advertisement may not "Include any untrue statement of a material fact" and may not "Include a material statement of fact that the adviser does not have a reasonable basis for believing it will be able to substantiate upon demand by the Commission".

This guide turns those prohibitions into seven steps for AI claims. Unless a step quotes a rule or an order, what it describes is our recommendation.

Step 1: Find every place the firm describes its AI

Delphia's statements appeared, according to its order, "in its Form ADV Part 2A brochures, in a press release, and on its website", and later in emails to new investors and a social media post. Global Predictions used its website, emails to clients and prospects, and social media. Inventory all of those channels, plus pitch decks, due diligence questionnaires, RFP responses, client chatbots and third-party content published for the firm.

The Marketing Rule defines an advertisement as a communication to more than one person that offers the adviser's advisory services to prospective clients, or new services to current clients, and it covers a communication "to one or more persons if the communication includes hypothetical performance". It excludes "Extemporaneous, live, oral communications" and information in a required filing reasonably designed to satisfy it. Brochure statements still matter: Delphia's order describes false statements in its Form ADV brochures as well as its advertisements, and found violations both of the Marketing Rule and of Section 206(2), which makes it unlawful to "engage in any transaction, practice, or course of business which operates as a fraud or deceit upon any client or prospective client." The order does not say which statements violated which provision. Treat every channel as in scope for accuracy, whichever rule applies (interpretation).

Step 2: Break each AI claim into checkable statements of fact

Phrases like "AI-powered" bundle several facts. Separate them: what the system does, which part of the service uses it, what data goes in, whose data it is, who built it, and any comparison, ranking or result.

Delphia's website said it "put[s] collective data to work to make our artificial intelligence smarter so it can predict which companies and trends are about to make it big and invest in them before everyone else." The order found that it collected some client data but "never used that data with artificial intelligence or machine learning or otherwise used that data in any way as inputs into its investing algorithms." Global Predictions claimed its technology incorporated "[e]xpert AI-driven forecasts," which the order found it did not, and called itself the "first regulated AI financial advisor", a claim it "could not produce documents to substantiate". Claims of being first or best deserve particular care, because they are facts about every other firm too (interpretation).

Be precise about which part is AI. Global Predictions' order notes that clients interacted through a chatbot, but "the chatbot does not generate allocation recommendations." If AI only drafts emails or summarizes research, marketing should not suggest it picks investments (interpretation).

Step 3: Build the substantiation file before the claim goes out

The rule asks whether the adviser had a reasonable basis, when it disseminated the advertisement, for believing it could substantiate the claim on demand. When it adopted the rule, the SEC said advisers "could make a record contemporaneous with the advertisement demonstrating the basis for their belief", and that "if an adviser is unable to substantiate the material claims of fact made in an advertisement when the Commission demands it, we will presume that the adviser did not have a reasonable basis for its belief." Both passages are quoted in the SEC examination staff's April 2024 risk alert on the rule, a staff document that, like all staff statements, says it has "no legal force or effect".

For each AI claim, the file should hold (interpretation): a description of the system from the people who run it; any vendor or model used; what data it uses; who checked the claim and when; and the evidence behind any comparison or result.

If an AI claim comes with model, backtested or projected results, the rule treats those as hypothetical performance (with an exclusion for certain interactive analysis tools), which may be advertised only if the adviser meets conditions, including policies and procedures making it relevant to the intended audience. Global Predictions' order found violations of that paragraph too, over model results for years predating the firm. Rule 204-2 separately requires records supporting any performance an advertisement presents.

Keep the substantiation file with it.

Step 4: State the limits alongside the benefits

The rule also says an advertisement may not "Include information that would reasonably be likely to cause an untrue or misleading implication or inference to be drawn concerning a material fact relating to the investment adviser", and may not discuss potential benefits "without providing fair and balanced treatment of any material risks or material limitations associated with the potential benefits".

For AI, that means saying what humans still decide, what the system cannot do and where outputs are checked (interpretation). A NASAA Investment Adviser Section publication, which says it does not necessarily represent NASAA's views, warns against "fake AI", such as services described as "powered by artificial intelligence" that rest on simple rules or manual processes. If a feature is rules-based, describe it as rules-based (interpretation).

Step 5: Keep AI claims current when the technology changes

A true claim can become untrue when a model, feature or data source changes. Delphia's order shows the risk of partial correction: after an SEC examination it updated its brochure and promised a marketing review, but statements about client data "continued to be made through August 2023" in emails, a social media post and a press release.

Give every AI claim an owner and a review date, re-check claims whenever the underlying system changes, and remove a withdrawn claim from every channel in Step 1, not just where the problem was found (interpretation).

Step 6: Write AI claims into the marketing review process

Rule 206(4)-7 requires SEC-registered advisers to adopt and implement written policies and procedures and to review them "no less frequently than annually". Both orders found failures here. Delphia's order says it "lacked policies and procedures relating to social media" and had no clear review and approval process for the employees and consultants involved. Global Predictions had policies requiring written approval of marketing before dissemination and a log of approvals, and the order says it "failed to implement any of these policies and procedures."

The examination staff's April 2024 risk alert also observed policies that "Did not address applicable marketing channels utilized by the advisers, such as websites and social media." In practice (interpretation): add AI claims to the marketing checklist, require the substantiation file before approval and keep the approval log the policy promises.

FINRA's advertising regulation FAQ tells member firms that "Firms are responsible for their communications, regardless of whether they are generated by a human or AI technology." The Marketing Rule has no exception for AI-drafted text, so the same review applies to advisers (interpretation).

Step 7: Prepare for the examiner's questions about AI

The SEC Division of Examinations' fiscal 2026 priorities, published November 2025, said the staff "will review for accuracy registrant representations regarding their AI capabilities or AI." The fiscal 2027 priorities, announced October 23, 2026, do not repeat that sentence. They keep language that also appeared in fiscal 2026: for firms offering automated investment advisory services and related tools, assessments "generally will include whether: (1) representations are fair and accurate". The edition's opening message says the Division is "intensifying its efforts to understand how the technologies influence outcomes and whether firms continue to comply with existing, technology-neutral obligations." Both editions say they are staff views with no legal force.

The fiscal 2027 edition uses a new term: "Registrants are increasingly incorporating super intelligence (SI) into their businesses." Its examples are tools firms already use, for "research, drafting, coding, and client communications, as well as advisory and portfolio management functions". No document reviewed for this guide defines SI, but the SEC's page on the subject places SI headings over documents still titled AI, so we read SI as the SEC's current term for what it previously called AI, not a reference to some future technology (interpretation). Marketing that adopts the term needs the same substantiation as any other claim (interpretation).

Checklist

• Every channel describing the firm's AI is inventoried, including third-party content
• Each AI claim is split into separate facts
• Each fact has a substantiation file before publication
• Any model, backtested or projected results meet the hypothetical performance conditions
• Claims of being first, best or unique are evidenced or removed
• Marketing says what the AI does not do and what humans still decide
• Advertisements and their substantiation are kept for the Rule 204-2 period
• Every AI claim has an owner and review date, and is re-checked when the system changes
• Withdrawn claims are removed from every channel
• Marketing policies cover AI claims in every channel, with an approval log
• AI-drafted marketing gets the same review
• Dual registrants also apply FINRA Rule 2210 to broker-dealer communications
• The inventory, files, approvals and re-check evidence are ready for an examiner

How we sourced this guide: every factual statement in it is drawn from published rules, regulatory guidance and other public documents, and we keep a full set of sources and citations for it. If you would like them, or want to question a point, get in touch through our contact page and we will share them.