The short answer
No new rules. The IRS wants your AI decisions written down.
The IRS office that can suspend or disbar tax practitioners has set out how their existing duties apply to work AI helped with. It tells them to use only secure, enterprise-approved AI for client data, and it never says what that term means. Practitioners who oversee the firm must take reasonable steps to make sure it has adequate procedures. For AI, the alert wants every step written down. So a partner should list which AI tools touch client information, who approved each one, and where that's written.
On June 24, 2026, the IRS Office of Professional Responsibility issued Alert 2026-19, Introductory Guidelines for Responsible AI Use in Federal Tax Practice. That office enforces Circular 230, and the alert sets out how a practitioner's existing duties apply to work AI helped produce.
Most of the coverage since has been about hallucinations, where AI makes things up: fake citations, sanctioned lawyers and a reminder to check your work. That part is real, and it's the part every firm already understood.
The part a partner should notice is quieter, and it sits in a different section. There the alert uses a word the IRS hasn't used about AI before, asking for a firm's steps to be documented.
The six rules the alert applies to AI
The alert doesn't create a new rule. It shows how six existing rules apply to AI, so a firm that met its duties on June 23 still met them on June 25. What changed is that a firm can no longer say it was unclear how the rules applied.
§ 10.22, due diligence. A practitioner has to check the facts, citations and calculations AI produced before the work reaches a client or the IRS. AI output is a draft.
§ 10.27(a), fees. Unconscionable fees are prohibited, and the alert applies that to billing for time that AI saved and nobody actually spent.
§ 10.35, competence. Competence now includes knowing how the AI tool produces its content and where it gets things wrong.
§ 10.36, procedures to ensure compliance. This is the duty that covers the firm's procedures as a whole, and we come back to it below.
§ 10.37, written advice. Advice has to rest on assumptions the practitioner checked independently. The alert notes that relying on a tool may not be reasonable if nobody can see how it reached its answer.
IRC §§ 6713 and 7216(a). These are the civil and criminal penalties for a preparer who discloses or uses return information without authorization. § 10.51(a)(15) ties them back into Circular 230. Under it, willfully disclosing or otherwise using a return or return information in a way the Internal Revenue Code doesn't authorize is a basis for discipline in its own right.
Of the six, the last pair matters most, because a disclosure to an AI tool the firm doesn't govern is more than a risk under the tax code. It's also a separate professional conduct matter, in front of the office that can suspend or disbar a practitioner from practice before the IRS.
The alert never defines enterprise-approved
On protecting client data, the alert is direct. It tells practitioners to handle client data using only secure, enterprise-approved AI with confidentiality safeguards in place, and never to upload sensitive data to unsecured sites.
The alert doesn't define that phrase. Circular 230 doesn't either, and no regulation currently on the books does, so your firm is held to a standard it can't look up. In practice, three readings are going around, and some are easier to defend than others.
The vendor reading: enterprise-approved means a paid tier. It's the weakest of the three. A business subscription changes the contract. By itself it doesn't change who can reach the data, where it's stored, or whether the firm can prove any of that later.
The tenancy reading: enterprise-approved means the data stays inside a boundary the firm controls. This one is stronger and closer to what the alert is worried about. On this reading, the tool runs in an environment the firm administers, under an agreement the firm holds. The company that provides the AI model sits outside that boundary.
The evidence reading: enterprise-approved means the firm approved it and wrote that down. Someone with authority looked at the tool, decided it was fit for client data and wrote down why. The firm can produce that decision when asked. Note that the alert's word is approved, which is something a person does rather than a feature a product has.
Read § 10.36 alongside the four things the alert says that section asks of a firm, which are set out below. The third reading then starts to look like the plain requirement.
Why the partners are responsible
Section 10.36 reaches past the staff member who pasted a K-1 into a chatbot, to the people who run the practice. In the regulation's terms, that's any individual subject to the provisions of Circular 230 who has, or shares, principal authority and responsibility for overseeing the firm's practice.
Any individual subject to the provisions of this part who has, or shares, principal authority and responsibility for overseeing a firm's practice governed by this part must take reasonable steps to ensure that the firm has adequate procedures in effect for all members, associates and employees, for purposes of complying with subparts A, B and C of this part, as applicable.
31 CFR Subtitle A, Part 10 § 10.36(a), condensedA single mistake doesn't bring discipline under § 10.36(b). Discipline attaches when the people in charge fail to put procedures in place, or fail to see that they're followed, through willfulness, recklessness or gross incompetence. The result also has to be a pattern or practice inside the firm.
The way AI gets into a small practice fits that pattern-or-practice test uncomfortably closely, because an unmanaged tool produces exactly that kind of pattern. It arrives as a habit. One person finds it useful in February and shows two colleagues in March. By the next filing season the firm has a practice that nobody decided on or wrote down, and that nobody owns or watches.
The alert then lists four things it says § 10.36 asks of a firm using AI.
Training. Staff training on the risks and the requirements.
Internal protocols. Ways of handling client data securely and of monitoring AI output for accuracy.
Vetting. Checking any third-party AI tool before the firm signs a contract for it.
Documentation. In the alert's own words, all steps and processes documented to show adherence to the section.
That last item is the reason for this brief. It asks a firm to be able to show its good practices as well as have them. A firm can keep every one of those procedures in someone's head, be genuinely careful, and still have nothing to hand the person who asks. An examiner can't give credit for a safeguard they can't see, and neither can an underwriter or a client's counsel during due diligence.
What the alert says about billing
The most surprising paragraph in the alert has nothing to do with security, and the commentary has almost entirely skipped it. Under § 10.27(a), a practitioner may not charge an unconscionable fee, and the alert applies that rule to AI directly. Billing for manual labor or time that was not actually spent may cross that line, depending on the facts. So may double billing for a task AI assisted.
The alert names two facts that bear on whether billing crosses that line: a noticeable pattern across clients, and the size of the billing differences. It says cost savings should be passed on openly: disclose what AI did, and credit the reductions to the client's account.
For a firm that bills by the hour, this risk is real and grows quietly. Every hour AI saves is an hour that must not appear on an invoice as though someone worked it. Nobody plans to bill it, but it happens anyway, because the timesheet habit is older than the tool.
Suppose your firm can show what AI did on an engagement and what it credited back. That puts it in a much better position than a firm relying on the fact that nobody has asked yet. And the same written record that answers § 10.36 answers the billing question too.
What your firm should be able to show
There are six written records, and none of them needs new software. A firm of ten to twenty-five people can put the set together in a few weeks.
An approved tools list, with the approver. It says which AI tools may touch client information. The decision belongs to someone whose role gives them that authority. Whoever happened to set the tool up doesn't get it by doing so. Record the date of the decision and what it rested on. This is the record that turns enterprise-approved from a description into a fact.
A written AI acceptable use policy. It sets out what staff may and may not enter into an AI tool, and what has to be reviewed before it leaves the firm. It also says what a staff member does when they're unsure. Keep it short enough that people will read it.
A dated training record. It shows who was trained, on what, and when. The alert names training first of its four items. This record is the cheapest of the six to produce.
A vendor assessment for each approved tool. It records where the data goes, whether it trains a public model, what the agreement says, and who at the firm accepted it. The alert asks for third-party tools to be vetted, and this is what vetting looks like once it's written down.
A review standard for AI-assisted work. It says who checks AI-assisted work before it carries the firm's name, and how that check is marked. §§ 10.22 and 10.37 both land here.
A § 7216 consent position, split by return type. This records whether the firm relies on a client's written consent or on an exception. For clients filing in the Form 1040 series, Rev. Proc. 2013-14 governs the format. The consent carries the mandatory language, and each separate disclosure or use gets its own written document. That document may be furnished as an attachment to an engagement letter, rather than as a clause inside one. For clients not filing in the 1040 series, the regulation is less prescriptive. A consent may be in any format, including in the text of an engagement letter, as long as it meets § 301.7216-3(a)(3)(i). Timing doesn't vary by return type: the consent is written, and it comes before the disclosure. If the firm is relying on an exception instead of a consent, the record is a written analysis of which exception and why.
Here are the sources for the paragraph above, since who each rule covers is the point most often gotten wrong. Rev. Proc. 2013-14 § 1 is the source for the Form 1040 series limit, and § 2.01 for the separate-document rule. The treatment of taxpayers outside the 1040 series comes from 26 CFR § 301.7216-3(a)(3).
There's no product or platform on the list, and no seal of approval either. None exists for this.
What to do in the next two weeks
Do this in the next two weeks, before the next filing deadline crowds it out. A partner should put three questions to the practice itself, even if an IT provider runs the firm's systems. The alert makes answering them a duty of the firm.
Which AI tools touch client information at this firm today?
Who approved them?
Where is that written?
In most firms of this size, the honest answer to the first question is longer than the partners expect. For the second and third, it's that nothing exists yet. That's a gap in the written record, and it's cheaper to close than a gap in security. Closing it starts with writing down decisions the firm has mostly already made.
Read the alert itself before someone quotes it back to you.
Related questions
Does OPR Alert 2026-19 create new rules for AI in tax practice?+
No. It shows how six obligations that already existed apply to AI-assisted work. Five are Circular 230 sections: 10.22 on due diligence, 10.27(a) on fees, 10.35 on competence, 10.36 on procedures to ensure compliance and 10.37 on written advice. The sixth is the preparer penalties at IRC 6713 and 7216(a), which section 10.51(a)(15) ties back into Circular 230. A firm that met its duties on June 23, 2026 still met them on June 25. What changed is that it can no longer argue that nobody knew how the existing duties applied.
What does enterprise-approved mean?+
The alert uses the phrase and doesn't define it. Circular 230 doesn't either, and no regulation currently on the books does. Three readings are going around. The first is that it means a paid vendor tier, and the second that the data stays inside a boundary the firm controls. The third is that someone with authority approved the tool for client data and wrote down why. That's the one that fits when you read section 10.36 alongside the four things the alert says that section asks of a firm. It's also the only reading that treats approving as something a person does.
Which section of Circular 230 reaches the partners?+
Section 10.36. It reaches any individual subject to the provisions of Circular 230 who has, or shares, principal authority and responsibility for overseeing the firm's practice. That person must take reasonable steps to ensure the firm has adequate procedures in effect for all members, associates and employees. A single mistake doesn't bring discipline under 10.36(b). Discipline applies when someone fails to put procedures in place, or fails to see they're followed, through willfulness, recklessness or gross incompetence. The result also has to be a pattern or practice inside the firm.
Does the alert say anything about billing?+
Yes, and it's the paragraph most often skipped. Under section 10.27(a), a practitioner may not charge an unconscionable fee. The alert applies that to billing for manual labor or time that was not actually spent, or double billing for a task AI assisted. The two facts it names are a noticeable pattern across clients and the size of the billing differences. It says cost savings should be passed on openly: disclose what AI did, and credit the reductions to the client's account.
What should a firm be able to produce?+
Six written records, none of which needs new software. The first three are an approved tools list naming the approver and the date of the decision, a written AI acceptable use policy and a dated training record. Then there's a vendor assessment for each approved tool, covering where the data goes and what the agreement says. The fifth is a review standard saying who checks AI-assisted work before it carries the firm's name. Last, a section 7216 position split by return type states which written consent or which exception the firm is relying on, and why.
Is there an official approval for enterprise-approved AI?+
No. There's no official approval for this, because no regulation defines the phrase. The firm's own written approval decision answers the question instead: which tools may touch client information, who decided that, on what date, and what the decision rested on.
Claremont Security helps regulated firms get more done with AI on client work, with the setup and the record to show for it, whichever tool the firm uses. Our AI Exposure Check runs the § 7216 question against your own tools, free and with no account: claremontsecurity.com/ai-exposure
Get a one-page estimate for your firm: what the work would take and what it’s worth, from your own numbers. No charge.
claremontsecurity.com/insights/enterprise-approved-ai-circular-230
