How to Disclose AI Use in Form ADV Part 2A: Brochure Language You Can Defend

1. The Short Answer: No Form ADV Item Names AI

No Form ADV item names AI. Part 1A, Part 2 and the General Instructions never mention artificial intelligence, machine learning or super intelligence.

The next deadline is March 31, 2027, when an adviser with a December 31 fiscal year end files its annual updating amendment. Before then, put AI in your Part 2A brochure wherever it is material to how you advise, describe it accurately in both directions, and keep every other document consistent with it.

Both directions matter. Claiming more AI than you use is what the SEC called AI washing in its 2024 cases against two advisers. Saying nothing about AI that shapes your advice leaves a gap in the description of how you advise.

This guide is operational guidance; your counsel and CCO own the interpretation. The inventory, the materiality questions and the wording advice below are WorkWise Solutions' view, and every fact links to its source.

The form never asks about AI. Clients and examiners read what you say about it anyway.

2. Why Brochures Matter More in 2027

The SEC's Division of Examinations released its FY2027 examination priorities on October 23, 2026. The document calls AI "super intelligence (SI)" throughout and opens the subject plainly: "Registrants are increasingly incorporating super intelligence (SI) into their businesses."

The standard is an old one: examiners will look at "whether firms continue to comply with existing, technology-neutral obligations" (p. 2). There is no new SI rule, and the guide to the 2027 priorities explains the term.

One paragraph matters most for a brochure. It covers automated investment tools and trading algorithms, without naming SI (p. 12), and says examiners will review whether "(1) representations are fair and accurate; (2) operations and controls are consistent with disclosures made to investors". An AI paragraph has to pass both halves.

The Division also plans to use SI itself, to "integrate SI responsibly into examinations and risk assessment" (p. 3).

SEC staff have already read brochures that way. In the custody proposal issued October 1, 2026 (Release No. IA-7023, File No. S7-2026-35), the economic analysis describes asking a large language model, named as Anthropic Sonnet 5, "to determine whether the adviser is giving or planning to give investment advice related to crypto assets." The review covered brochures from 15,964 advisers.

The release reports "Based on an AI-assisted review of these disclosures, we identified 1,498 advisers" and adds "This process found a false positive rate below 3%." Its request for comment 340 asks the public: "Are there limitations or potential sources of error in using a large language model for this purpose?" Comments are due by December 7, 2026.

Be exact about what the model did. It classified crypto advice, and nothing in the release says it looked for AI use. What it shows is that a model can now read nearly 16,000 advisers' brochures for a single question.

Write the AI paragraph as if a machine will read it beside every other adviser's. Brochures have already been read that way once.

3. What the SEC Has Charged So Far

On March 18, 2024, the SEC announced charges against two advisers, Delphia (USA) and Global Predictions, for false and misleading statements about their use of AI, including in Form ADV Part 2A brochures. The penalties were $225,000 and $175,000, $400,000 in total, under the Marketing Rule and the compliance rule.

Gary Gensler's comment in that announcement, quoted below, is still the shortest statement of the standard.

Rimar Capital came next, in an SEC order of October 10, 2024. The SEC's April 30, 2026 distribution-plan order describes that case as involving "an adviser that purported to use artificial intelligence to perform automated trading for advisory client accounts."

Since then the record for advisers is thin. A search for this guide on October 28, 2026 found no 2025 or 2026 SEC action announced against a registered adviser for false claims about AI. The nearby cases involve others: the founder of Nate, Inc., an issuer, charged over false statements about the company's use of AI (April 7, 2026), and fraud schemes built on supposed "AI-generated trading signals" (September 29, 2026).

No AI-specific rule sits behind any of these cases, and none is coming from the 2023 proposal. The SEC's predictive data analytics proposal was withdrawn on June 17, 2025, with the words "The Commission does not intend to issue final rules with respect to these proposals."

Each case turns on the same gap: what a firm said its AI did, against what it actually did.

4. Start With an Inventory

You can't describe what you haven't listed. In WorkWise Solutions' view, the brochure work starts with a one-page inventory of every AI tool the firm uses, finished before anyone drafts a sentence.

The 2027 priorities supply the categories. Registrants, the document says (p. 2), are "using generative SI tools to support research, drafting, coding, and client communications, as well as advisory and portfolio management functions, and compliance monitoring and surveillance." Sort every tool into one of those.

For each tool, record six things:

• Tool and plan. The product, the vendor and the plan the firm pays for.
• Use. Which category it serves, and for which clients.
• People. Who uses it and who checks its output.
• Client data. Whether client information goes in, and on what terms.
• Influence. Whether it shapes what you recommend, buy, sell, charge or review.
• Public description. Every place the firm describes it: brochure, website, pitch book, DDQ answers and client reports.

Include tools staff use without approval; the shadow AI guide covers how to find them. Ask research and model providers whether AI shapes the output you rely on, and let counsel decide whether that reaches your brochure.

The last column finds the problems. A tool praised in the pitch book and missing from the inventory is a claim with nothing behind it.

5. When AI Is Material Enough for the Brochure

Materiality is the firm's call, made with counsel. Two examples mark the ends of the range.

An analyst uses AI to draft an internal email or tidy meeting notes. The work gets faster, and what clients receive and pay stays the same.

A model screens securities, ranks ideas or proposes allocations that portfolio managers act on. That shapes the advice itself.

Most tools sit between those ends. In WorkWise Solutions' view, five questions sort them:

1. Does it influence which securities or strategies you recommend? Look at Item 8.
2. Does it change what clients pay? Look at Item 5.
3. Does it create a conflict, such as a stake in the vendor or payment from it? Look at Items 10 and 11.
4. Does it decide which accounts get reviewed, or when? Look at Item 13.
5. Is it part of the service clients buy, visible in their reports, plans or messages? Look at Item 4.

A yes to any question points to a brochure item. Five noes usually mean the tool belongs in your policies and inventory, where examiners reviewing "policies and procedures relating to monitoring SI" (p. 12) can still find it.

A tool earns a brochure sentence by changing the advice, the fee, a conflict or a review.

6. Item by Item: Where AI Can Show Up in Part 2A

Part 2A has no AI item, so AI goes wherever the facts of its use belong. The Part 2 instructions name the items, and the table maps AI onto six of them.

Item 2 comes last because you write it last. Once AI changes another item materially, the summary of material changes should say so in plain words.

Resist a standalone section headed Our Use of AI. It tends to collect marketing language, and it separates the AI from the method, fee or review it belongs to.

Claims of a new AI item are false. The current form, SEC 1707 (07-24), has none, and Part 1A's Item 5.L, titled Marketing Activities, says nothing about AI either.

Put each fact where the form already asks for it.

7. Describing AI Risks Without Boilerplate

Item 8's full name ends in Risk of Loss, and that is where AI risk language belongs when AI is part of your method. The easy version is generic warnings that could sit in any brochure.

In WorkWise Solutions' view, four habits make the paragraph yours:

• Tie each risk to a tool you use. If no model selects securities at your firm, a risk factor about one describes some other firm.
• Name the failure. A summary that misstates a filing, or a vendor that changes its model without notice.
• Pair each risk with its control. Who checks the output, how often, and what happens when the tool is wrong.
• Use plain words. A generative AI tool that summarizes filings tells a client more than a proprietary intelligence engine does.

Here is the difference, in illustrative wording to adapt with counsel:

• Vague: We use advanced artificial intelligence to optimize portfolios and reduce risk.
• Specific: Our analysts use a generative AI tool to summarize company filings and research reports. An analyst checks each summary against the source document before it informs a recommendation. Portfolio managers make all investment decisions. The tool can misread or omit information, which is why every summary is checked.

The vague version promises an outcome nobody can test. The specific one names the tool's job, the check and the limit.

If a sentence can't be tested against what the firm does, cut it.

8. Overstatement and Understatement

Accuracy runs both ways, and a study published October 14, 2026 in Frontiers in Artificial Intelligence measures what advisers actually write.

Samir Chincholikar and Robin Chawla sampled 400 of 16,223 advisers and classified their brochures with GPT-4o, another large language model put to work on Form ADV. Their headline: "AI use is disclosed in 23.7% of brochures (22.7% survey-weighted), while 27.6% contain AI-related risk language."

So more brochures warn about AI risk than describe AI use. A risk factor with no matching description leaves a reader asking which is true: does the firm use AI or not?

The words matter too. The authors report that restricting to the literal phrase "artificial intelligence" and weighting to the wealth and retail population "gives 9.9%". So when you check your own brochure, search for every term your firm uses: AI, machine learning, model, algorithm, automated.

Overstatement has a record: the 2024 cases. Understatement has a cost too. If a model shapes your advice and the brochure is silent, an examiner checking whether operations are consistent with disclosures made to investors can find the gap.

Understate, and the brochure is incomplete. Overstate, and the brochure becomes the evidence.

9. One Source of Truth Across Every Document

The brochure is one of several places your firm describes its AI. The 2024 cases ran under the Marketing Rule as well as the compliance rule, so the website and pitch book need the same care as Part 2A.

In WorkWise Solutions' view, the fix is one approved description per tool, kept in the inventory and reused everywhere:

• The brochure takes the relevant sentences, item by item.
• The website and pitch book start from the same text, and anything new goes through the same review.
• DDQ and RFP answers quote the inventory. The guide on what to tell LPs about AI covers the questions investors ask.
• Client reports and letters describe each tool the way the brochure does.

When a tool changes, its description changes in every document on the same day, and a change log records when. The guide to AI washing in LP materials has a before-and-after rewrite table for investor decks.

Five documents that say the same thing are a disclosure. Five that differ hand an examiner the first question.

10. The Calendar: March 31 and April 30, 2027

General Instruction 4 sets the first date: "You must amend your Form ADV each year by filing an annual updating amendment within 90 days after the end of your fiscal year." For a December 31 fiscal year end, that is March 31, 2027.

Clients come next. Under rule 204-3(b), the Part 2 instructions require you to "deliver, within 120 days of the end of your fiscal year, to each client a free updated brochure that either includes a summary of material changes or is accompanied by a summary of material changes". For a December 31 year end, that is April 30, 2027.

Annual is the minimum. Between annual filings, the brochure has to be amended when information in it becomes materially inaccurate, so a material change in how you use AI mid-year doesn't wait for March. Counsel decides when a change is material.

A working timeline for a December 31 year end, in WorkWise Solutions' view:

1. November and December: build the inventory and settle materiality with the CCO.
2. January: draft the AI language item by item, plus the matching website, pitch book and DDQ text.
3. February: counsel reviews, and each claim is tested against the inventory with its evidence filed.
4. By March 31: file the annual updating amendment.
5. By April 30: deliver the brochure with its summary of material changes.

Work backward from March 31. The inventory takes longest, so it starts first.

11. Where to Start

This week, open your current brochure and search it.

1. Search every term your firm uses for AI in the brochure, website, pitch book and last DDQ, and paste each sentence into one document.
2. List every AI tool, approved or not, with the six columns from section 4.
3. Run the five materiality questions with your CCO.
4. Rewrite each claim to match what the tool does, with the evidence beside it.
5. Put March 31 and April 30, 2027 in the calendar, with counsel's review booked before both.

The 2027 exam priorities guide covers the rest of the AI file examiners may ask for, and the deepfake fraud guide covers the fraud controls.