The IRS Just Made AI Billing an Ethics Issue — What Every CPA Firm Needs to Do This Month

In June 2026, the IRS Office of Professional Responsibility issued Alert 2026-19. It did not create new rules. It applied existing ones to a behavior a lot of accounting firms have been practicing quietly: using AI to cut delivery time, then billing clients as if nothing changed.

That behavior now has a name under Circular 230. The OPR calls it a potential unconscionable fee.

If your firm has used AI tools to speed up tax research, return preparation, or document review in 2026 — and your billing has not changed — this alert is about you.

What the IRS OPR Actually Said

Alert 2026-19 addresses two Circular 230 provisions. Both already existed. The OPR's contribution was to apply them explicitly to AI-assisted practice.

The Unconscionable Fee Problem

Circular 230 prohibits fees that are "unconscionable." The OPR's June guidance applied this to AI efficiency gains. The plain-language version: if AI reduces an eight-hour task to two hours, billing the client for eight hours of work may violate Circular 230.

The specific language: practitioners must "fairly credit to the client's account any cost reductions" resulting from AI use.

That phrase — "fairly credit" — is deliberately open-ended. But the direction is clear. AI efficiency is supposed to flow to the client in some form, not sit entirely inside the firm's margin. If a client is paying for 40 hours of return preparation and your team now delivers it in 8 hours because of AI tools, your billing needs to be justifiable. "That is what I used to charge" is not a defense in a disciplinary proceeding.

This does not mean your rates must decrease. It means they must be explainable. Flat-fee pricing, value-based pricing, and explicit disclosure are all defensible. Billing time you did not spend is not.

The Confidentiality Problem

The OPR also applied Circular 230's duty of confidentiality to AI tool selection. Uploading client personal or financial data to public AI tools — the free tier of any general-purpose AI, uncontracted third-party apps — without client consent is now explicitly framed as a Circular 230 compliance issue, not just a data risk.

The standard: AI tools that process client data must be covered under a written data processing agreement, or clients must have explicitly consented to the data being processed by that tool.

For most small CPA firms, this means: if you are running client financials through any AI tool to speed up analysis, that tool needs a signed contract with your firm. Free tier means no contract means not covered.

Enterprise-tier tools from Intuit, Thomson Reuters, Wolters Kluwer, and comparable vendors typically come with data agreements. Those are fine. Public AI tools used ad hoc for client data are the specific risk.

The Verification Standard

A third implication of Alert 2026-19 is often overlooked. The OPR requires practitioners to verify every AI-generated document before it reaches the IRS or a client. Not most documents. Every document. "The AI drafted it" is not a defense for a filing error. The standard of professional care applies to the output, regardless of how it was produced.

This is not new — practitioners have always been responsible for their filings. What is new is that some firms have been treating AI output as finished work rather than a first draft. That posture is now explicitly inconsistent with the guidance.

What "Fairly Credit" Means in Practice

The OPR did not prescribe a pricing structure. It told you what is no longer defensible. There are three viable paths:

Flat-fee repricing for commoditized services. If AI has made you efficient enough at a specific service — individual returns, bookkeeping maintenance, routine compliance memos — that hourly billing no longer maps to the value delivered, move to a flat fee. Price based on what the service is worth to the client, not how long it takes. This is a better model for both parties: clients get predictability, you get scalability. The "fairly credit" standard is automatically satisfied because time is not the pricing unit.

Billing actual time for time-based engagements. If you keep time-based billing, bill actual time. AI-augmented work that takes two hours should be billed as two hours. Track it honestly. The record-keeping discipline that was always part of good practice is now also a compliance backstop.

Disclosure in the engagement letter. If you want to maintain your current fee structure while using AI, the defensible path is informed consent. Your engagement letter states how AI is used, what it affects, and how that shapes your pricing. Clients who sign that agreement have consented. You have documentation. That combination protects you under both the fee and confidentiality provisions.

Most small CPA firms land on a hybrid: flat fees for the commodity work that AI has fully automated, updated engagement letters for everything else. The Bronze/Silver/Gold subscription model is the most practical template for structuring those tiers.

The Two Actions to Take This Month

Action 1: Review AI-assisted engagements billed in 2026

Go back through your billing records for this year. Identify engagements where AI tools materially reduced delivery time — tax return preparation, document analysis, research memos. For each one, ask: if a client or a disciplinary committee asked you to justify that bill, what would you say?

You do not need to proactively issue refunds or disclosures for past work. You need to know your exposure and decide how to handle it going forward. If you identify clear cases where billed time and actual time diverge significantly, talk to your malpractice insurer. Document that conversation.

The purpose of this review is not to audit yourself retroactively into a problem — it is to understand your current posture so you can correct it going forward with full awareness.

Action 2: Update your engagement letter AI disclosure

Every new engagement letter going forward should include a one-paragraph AI disclosure that covers four things:

1. A statement that AI tools are used in the firm's service delivery
2. The categories of tasks where AI is applied (document review, research, return preparation, client communication drafting)
3. Your data handling policy — which types of tools the firm uses and that they are covered by written data agreements
4. A brief note on how AI efficiency factors into your billing approach — flat fees for automated work, time-based for complex judgment work, or another model

This paragraph creates the informed consent that protects you under both the unconscionable fee and confidentiality provisions.

If your engagement letter template is more than six months old, it predates this guidance. Update it before your next engagement starts.

The Larger Pattern This Alert Is Part Of

The OPR guidance lands at a moment when accounting firms are automating faster than they are updating their client relationships. The Journal of Accountancy's August 2026 reporting shows some small CPA firms have automated more than 80% of individual tax return preparation. That efficiency is real and valuable. The problem is when the efficiency gain sits entirely inside the firm — reduced hours, same invoice — without any corresponding update to billing structure or client disclosure.

The OPR alert is not hostile to AI adoption. Every source covering Alert 2026-19 notes that the guidance explicitly encourages AI use and does not suggest practitioners should avoid it. What it does require is that the adoption is matched by professional responsibility practices that have kept pace with the technology.

Firms that get ahead of it — with flat-fee repricing for commoditized services, updated engagement letters, and documented AI verification workflows — are on the right side of this shift. Firms that do not are running a compliance exposure that grows with every AI-assisted engagement billed at old rates.

The guidance has been out since June. The window to get ahead of it is now.