AI Regulation Tracker / Regulator guidance
IRS OPR Alert 2026-19 Puts Circular 230 Duties on AI Use in Tax Practice
The IRS Office of Professional Responsibility told CPAs, enrolled agents, and tax attorneys they must independently verify anything generative AI produces. Blind reliance on an AI tool can breach Circular 230.
The IRS Office of Professional Responsibility has put tax professionals on notice that using generative AI does not lower the bar set by Circular 230. In Alert Issue No. 2026-19, released June 24, 2026 and titled "Introductory Guidelines for Responsible AI Use in Federal Tax Practice," the office that polices practitioner conduct spelled out what responsible AI use looks like for anyone who practices before the IRS.
The message is direct. A practitioner who leans on an AI tool without checking its work is still fully responsible for the result. The alert frames AI output as a starting point, not an answer, and reminds practitioners that the professional standards they already know do not bend because a machine drafted the first version.
Verify everything the tool produces
The core instruction is independent authentication. According to the alert, practitioners must independently authenticate all factual and legal information generated by AI. Legal citations must be checked, cases must be read, and financial data must be confirmed before it reaches a client or a filing. The office cautions that "blind reliance on what AI yields, especially when the underlying logic or sources are unclear, may constitute unreasonable reliance."
That phrase is load-bearing. It ties AI use to Circular 230 section 10.37, which governs written advice and requires a practitioner's conclusions to rest on reasonable assumptions and reasonable reliance. The alert's position is that reliance on an AI projection or representation the practitioner cannot trace or explain is not reasonable. If the tool cites a case, the practitioner reads the case. If the tool produces a number, the practitioner confirms the number. The duty to get it right stays with the human.
Client data is the second exposure
The alert's second warning is about confidentiality. Tax return information is protected, and the office points to IRC section 7216 and section 6713, which impose criminal and civil penalties for the unauthorized disclosure or use of return information. The alert flags that generative AI platforms can create disclosure risk, "especially when data is uploaded to unsecured or public systems."
The practical guidance is to use only secure, enterprise-approved AI with strong confidentiality safeguards. For a working practitioner, that means the free public chatbot is not a place to paste a client's return, a W-2, or the facts of a matter. Section 7216 exposure attaches to the disclosure itself, so the moment protected information enters an unapproved tool, the risk is live regardless of what the tool then does with it.
What the alert does not do
This is guidance, not a new rule or regulation. The alert does not create fresh statutory duties. It operationalizes duties that already exist under Circular 230 and the Internal Revenue Code and applies them to a new tool. It does not ban generative AI in tax practice, and it does not set a certification or approval process for specific products. The office describes the guidelines as introductory, which signals more detail may follow rather than a finished rulebook.
It is also distinct from a general Circular 230 refresher. The value here is specificity: the alert names the AI failure modes, hallucinated citations, unverifiable projections, and careless data uploads, and maps each to an existing duty. It tells practitioners what to actually do now, not merely that the rules still apply.
Why it lands now
Regulators across professions have spent the past year responding to AI-fabricated citations and mishandled client data. Courts have sanctioned lawyers for filing hallucinated cases, and licensing bodies have issued parallel warnings. The IRS alert brings the tax bar into that pattern from the enforcement side, because the Office of Professional Responsibility is the body that can pursue discipline against practitioners who fall short of Circular 230. A practitioner reading this should treat it as the standard against which future conduct will be measured.
Frequently Asked Questions
What changed with IRS OPR Alert 2026-19?
The IRS Office of Professional Responsibility issued guidance on June 24, 2026 applying Circular 230 duties to generative AI. Practitioners must independently verify all AI-generated facts, citations, and financial figures, and blind reliance on opaque AI can count as unreasonable reliance under section 10.37.
Who does this affect?
Any tax practitioner who practices before the IRS, meaning CPAs, enrolled agents, and attorneys governed by Circular 230, along with the firms that supervise their AI use.
Does the alert ban using AI in tax practice?
No. It permits AI use but requires independent verification of output and restricts client return information to secure, enterprise-approved tools. It treats AI as a reviewable starting point, not a source you can rely on unchecked.
What is the penalty risk for putting client data into a public AI tool?
Unauthorized disclosure or use of tax return information can trigger penalties under IRC section 7216 and section 6713. The alert warns that uploading sensitive taxpayer data to unsecured or public AI systems creates that exposure.
Sponsored Training
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.