Can an accounting firm license its data to AI buyers?
· Company Data Advisors
Accounting and finance is the industry AI data buyers name most, but an AICPA ethics article advises declining unless it is clearly permitted. What to check first.
Maybe, but a CPA firm cannot answer it by asking what buyers want, because buyers want plenty. As of November 1, accounting and finance was the industry named by the most programs in our buyer index, 9 of 17, according to our November demand map. The harder question is whether the firm has the right to license what it holds. On November 2, 2026, the Journal of Accountancy, the AICPA's magazine, published an article by the communications manager of the AICPA Professional Ethics Division headlined "Asked to sell firm data for AI training? Be very cautious". It says the prudent default is to decline unless you can clearly establish that the transaction is permissible.
This is a reading of public sources, not legal or ethics advice. Have counsel review before you share anything with a buyer, including a sample.
What do AI data buyers want from accounting and finance firms?
AI data buyers want the operating records that show finance work being done: books, invoices, billing and ERP exports alongside the email and documents around them. Mode names accounting, finance and medical practices as a group and sets a lower employee minimum for them than for other businesses. Its source list includes QuickBooks, NetSuite and Paychex. The table shows eight of the programs that name accounting or finance records or firms.
| Buyer | What it names | Size or limit it states |
|---|---|---|
| Mode | Accounting, finance and medical practices; QuickBooks, NetSuite and Paychex among its sources | 10 or more employees |
| Mercor | An accounting listing among its 25 | US only for accounting |
| Turing | Finance and accounting among its industries | Not stated |
| Centific | Finance and accounting first among its industries | Not stated |
| Nyne | Finance, customer and scheduling records; names accounting systems | 20 or more full-time US W-2 employees |
| Polyshares | Accounting and ERP exports among its data types | No minimum |
| Appen | ERP and finance, HR and payroll, procurement records | Not stated |
| Handshake | Finance and insurance, and professional services | 20 or more full-time employees and 3 or more years in operation |
Sources for each row are on the buyer's page in our index, which links to the program's own page. Two details are easy to miss. Mercor's accounting listing is limited to US firms, according to its buyer page, while most other listings are open outside the US and a few prefer US firms. And Nyne is one of the few programs that names accounting systems explicitly.
What do the professional rules say?
The AICPA's Confidential Client Information Rule (ET sec. 1.700.001) generally prohibits members in public practice from disclosing confidential client information without the client's specific consent, subject to the rule's exceptions. That is the Journal of Accountancy's summary, and it adds four points that matter for a data license:
- Engagement consent is not enough. Permission to process information to perform an engagement "does not authorize you to hand the same information to a buyer (or any other party) to train an AI model."
- A friendly label does not change it. Calling the recipient a "technology partner" does not resolve the issue, because sharing data for the recipient's independent commercial purposes warrants different scrutiny than hiring a vendor to help deliver client services.
- Redaction does not turn client information into nonconfidential information. The article says not to treat redaction, or a buyer's "anonymous data" label, as permission.
- Consent has to say what is happening. It should explain the information involved, the recipient and the training purpose, not be buried in a generic technology disclosure.
The article is explicit about its limits: "The recommendation is caution, not a claim that AICPA rules categorically prohibit every data transfer." It also says AICPA ethics staff plan to publish a deeper dive later this quarter, so more detailed guidance is coming.
What does the tax-preparer law add?
Federal law adds a separate rule for tax return information. Section 7216 of the Internal Revenue Code makes it a misdemeanor for a person in the business of preparing returns or providing services in connection with preparing them, or for a paid preparer, to knowingly or recklessly disclose information furnished for preparing a return, or use it for any purpose other than preparing the return. The penalty is a fine, up to one year in prison, or both.
The Treasury regulation on consent, 26 CFR 301.7216-3, says that unless the statute or a related regulation authorizes a disclosure, a preparer may not disclose or use tax return information before getting the taxpayer's written consent. Among its requirements:
- The consent must be knowing and voluntary, and conditioning services on the taxpayer's consent makes it involuntary, with one narrow exception for disclosures to another preparer who helps prepare the return.
- It must name the preparer and the taxpayer, state the purpose, specify the information, and be signed and dated. Except for certain non-individual taxpayers, it must also identify the specific recipient.
- It cannot be given after the fact.
- If it does not state a duration, it is effective for one year from signing.
The Journal of Accountancy cautions not to assume that "an ordinary engagement letter is sufficient" and says to have counsel evaluate the transaction. A 2017 Tax Adviser column by two members of the AICPA Tax Practice Responsibilities Committee notes that the statute reaches reckless disclosure, not only intentional disclosure.
Which of an accounting firm's records are even candidates?
Our view: an accounting firm holds two different bodies of records, and only one of them is plausibly a starting point. The first is the firm's own operating history: how it bills, books its own time, runs its own finance and runs the practice. The second is client work: workpapers, returns, client email and anything that contains client information. The rules above mostly concern the second.
| Records | Our read | Why it matters |
|---|---|---|
| The firm's own books, billing and expense records | Candidate, after review | Firm information, but invoices and ledgers can name clients |
| Employee and payroll records | Usually leave out, or ask counsel | Personal data about staff; see our guide to employee and customer data |
| Client workpapers, returns and tax files | Do not include without specific client consent | The AICPA rule applies, and Section 7216 can apply if a file holds tax return information |
| Email, chat and shared drives | Treat as mixed | Firm business and client information sit in the same threads |
This is a screening view, not a conclusion about any firm. The Journal of Accountancy's reasoning, that redaction does not remove the duty, is why the mixed category belongs on the cautious side.
Not every firm that does finance work is a CPA firm. A bookkeeping company, a lender or a finance team inside an operating business is not bound by AICPA rules unless its people are AICPA members, though Section 7216 and state rules can still reach anyone who prepares returns, and its records still hold customers' and vendors' information and carry contract terms. Our guides to employee and customer data and to Slack, Google and Microsoft terms cover those questions.
What should a firm check before it applies?
Before applying, an accounting or finance firm should settle who owns the records, who must consent and what the buyer will do with them.
- Sort the records by who they belong to. Firm records, client records and mixed records need different answers.
- Read your engagement letters and privacy notices for what clients agreed to. The Journal of Accountancy says to examine client agreements and privacy obligations before considering a sale.
- Decide whether consent is realistic. If specific, written consent from each affected client is needed, ask whether you would request it and what that does to the relationship.
- Ask where de-identification happens. Mode says it removes sensitive material and transforms personal identifiers in an agreed copy before sale. Buyers differ on who does this and when. See our buyer-by-buyer comparison of license terms.
- Ask what the buyer can do onward and what happens to a trained model. The Journal of Accountancy asks whether the recipient can meaningfully restrict downstream use and address information already in a trained model.
- Get counsel involved first. The article's recommended gate is legal, ethics and information-security review. Our own suggestion: also ask your professional liability carrier.
Law firms face a parallel set of duties under the professional-conduct rules for lawyers; see our guide for law firms.
Where we come in
Company Data Advisors is an independent advisory firm. We track the published terms of every major program, have direct contacts at the major data buyers and at AI labs, and compare offers on the whole deal: price, exclusivity, how long the license runs, deletion and onward use. For an accounting firm, we can say which programs state limits for accounting and where each says de-identification happens. We cannot tell a firm whether its client consents are sufficient. That is a question for the firm's counsel.
Nothing about your company is shared until you approve a specific, named buyer. Participating buyers may pay us referral fees, and we tell you whether a buyer pays us before any introduction. Owners pay us nothing, and you can apply to any buyer directly.
What to do next
Start with the inventory, not the buyers. List the systems your firm runs on, mark each as firm, client or mixed, and take the list to counsel with the Journal of Accountancy article. If a clean category remains, answer eight quick questions to see which programs it may fit, or read our guide on how to sell or license your company's data to AI companies for the steps from first form to payment.
Programs in this post
More from the blog
- Is my company too small to license its data to AI buyers?
Most AI data buyers with a size floor set it at 20 to 50 employees, but law, accounting and medical firms qualify smaller, and nine programs state no minimum.
- What company data AI buyers want: November 2026
Across the 17 AI data-buyer programs we track, accounting, healthcare and software lead the industries, and CRM, ticket, email and finance records lead.
- How the company data market works: who buys and who pays
AI labs rarely buy company records directly; data companies license them and build training material. Who pays whom, the three kinds of deal, and the timeline.
See which buyers fit your company
Eight quick questions against every program in our buyer index. Participating buyers may pay us referral fees.