IRS AI alert leaves open whether sending return data to an AI tool is a Section 7216 disclosure
IRS Office of Professional Responsibility Alert 2026-19 tells tax practitioners to handle client data using only secure, enterprise-approved AI. CPA Steven de la Fe expects staff to paste part of a client's return into a chat tool this busy season.
The Investor · Invest desk
What happened
- On June 24 the IRS Office of Professional Responsibility issued the alert as its introductory guidelines for responsible AI use in federal tax practice.
- The alert warns against uploading sensitive information to unsecured or public systems and ties that to civil and criminal penalties for unauthorized disclosure under IRC Sections 6713 and 7216.
- Firm leaders with oversight duties under Circular 230 Section 10.36 need procedures for staff training, secure data handling, AI accuracy monitoring and vetting of outside tools, each step documented.
- The guidance also lists cross-client leakage among the privacy risks, meaning data from one client's file turns up in the answer to a question about a different client.
- Section 7216 requires written taxpayer consent before return information is used or disclosed outside the regulatory exceptions, and Rev. Proc. 2013-14 sets a separate signed format for Form 1040-series returns.
Why it matters
- decision Firm owners must choose before the busy season whether to collect a standalone signed AI consent from clients now or wait for an IRS answer that may not arrive in time.
- cost Because conditioning service on consent generally makes it involuntary, a firm that adopts AI also keeps staff able to prepare returns by hand, and the AI time saving applies only to clients who sign.
- exposure Approving a vendor does not close the gap if staff open the wrong plan: consumer and business tiers can carry different terms, and a zero-retention agreement may reach only the developer API.
Most of the alert, according to CPA Steven de la Fe, applies Circular 230 duties practitioners already know [13]. The penalties it points to are the statutory ones for unauthorized disclosure [3], and his column does not describe a separate penalty for a firm that has no written AI policy. A policy is evidence of the supervision Section 10.36 asks for [4]. It does not settle whether a given upload is a disclosure at all.
On that second question the IRS has not addressed whether giving return information to an AI tool amounts to a third-party disclosure [7]. De la Fe puts the test for owners this way: "The question for firm owners is whether the firm has decided, in writing, where that data is allowed to go." [16] We think that is the right test for the Section 10.36 duty. For Section 7216 it is incomplete. A written policy cannot tell a firm whether its approved tool needs the client's signature.
The IRS can answer in either direction, or not at all. A ruling that enterprise tools are not third-party disclosures would confine the consent problem to firms using public tools. A ruling that they are would make signed consent the price of every AI-assisted 1040-series return [6]. We think silence through the busy season is the likeliest of the three. That leaves firms with the standalone signed consent, worded with counsel, that de la Fe recommends until the IRS speaks [8]. If the IRS clears enterprise tools, that consent file was avoidable work and we were too cautious.
The log the alert asks for also matters for fees. De la Fe says a firm should be able to name, for any return, the tools that processed the client's information and the person who reviewed the output. If AI cut the time on a return, he says, the log shows it [14]. Under Section 10.27, the alert says, billing for manual time that was not spent may be a violation depending on the facts. So may double billing for AI-assisted work [5]. De la Fe reports that on the agent workflows he and a co-founder built for a high-volume practice, human time "came down to less than 5 minutes per return" [15]. If a firm's own log showed that, a bill built on manual hours would have to explain the gap.
What to watch
- Any IRS statement on whether sending return information to an AI tool is a third-party disclosure under Section 7216.
- The first enforcement action or OPR referral that cites IRC 6713, 7216 or Section 10.36 in connection with an AI upload.
- Written vendor terms confirming no training on client data and no retention for the exact plan a firm's staff use.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence50
- Adoption
- Insufficient
- Hype gap+10
- Incentives45
- Confidence55
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
On June 24, the IRS Office of Professional Responsibility issued OPR Alert 2026-19, its introductory guidelines for responsible AI use in federal tax practice.
- [2]
The alert says practitioners must handle client data "using only secure, enterprise-approved AI".
- [3]
The alert warns against uploading sensitive information to unsecured or public systems and ties that to the civil and criminal penalties for unauthorized disclosure under IRC Sections 6713 and 7216.
- [4]
The alert says firm leaders with oversight duties under Circular 230 Section 10.36 need procedures covering staff training, secure data handling, AI accuracy monitoring and vetting of outside AI tools, with every step documented.
- [5]
Under Section 10.27 on fees, the alert says billing for manual time that was not spent, or double billing for AI-assisted work, may be a violation depending on the facts.
- [6]
Section 7216 requires a taxpayer's written consent before a preparer uses or discloses return information outside the regulatory exceptions; for Form 1040-series returns, Rev. Proc. 2013-14 sets out a separate, signed consent format.
- [7]
The IRS has not said whether sending return information to an AI tool counts as a disclosure to a third party.
- [8]
Until the IRS says whether sending return information to an AI tool is a third-party disclosure, de la Fe recommends a standalone, signed AI consent, worded with your counsel.
- [9]
De la Fe writes that the regulations say a consent must be knowing and voluntary, and that conditioning services on it generally makes it involuntary.
- [10]
De la Fe says a firm needs a way to prepare a return without AI for any client who declines the consent, and staff need to know which clients those are before they open the file.
- [11]
The guidance names cross-client leakage as a privacy risk, where data from one client's file is reused to answer a question about another.
- [12]
De la Fe notes that consumer and business tiers of the same AI product can carry different terms, and asks whether a zero-retention agreement covers the chat app staff actually open or only the developer API.
- [13]
De la Fe writes that most of the alert applies Circular 230 duties practitioners already know.
- [14]
De la Fe says that for any return a firm should be able to answer which tools processed the client's information and who reviewed the output before it went out, and that if AI cut the time on a return the log shows it.
- [15]
De la Fe wrote that when he and his co-founder built agent workflows for a high-volume tax practice, "human time came down to less than 5 minutes per return."
- [16]
"The question for firm owners is whether the firm has decided, in writing, where that data is allowed to go."
- [17]
De la Fe writes that at a lot of firms this coming busy season, someone on staff will paste part of a client's return into a chat tool to get a quick summary, usually to clear a backlog.
Sources
1 independent publisher whose own reporting we read for this story.
- cpapracticeadvisor.comWhat to Settle Before Client Tax Data Touches AI
1 article · October 12, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.