Not a Substitute for Professional Judgment: The IRS Alert on AI in Tax Practice of June 2026

On June 24, 2026, the IRS sent out Issue Number 2026-19 in its alerts from the Office of Professional Responsibility. Its title is "Introductory Guidelines for Responsible AI Use in Federal Tax Practice." It takes provisions of Circular 230 that it quotes, and reads each against AI.

"Virtually all professional tax firms use some form of AI, whether they are aware of it or not," the alert says. It calls generative AI "GAI," meaning programs that "generate original content." Of generative AI it says that "it has limitations," among them "fabricated outputs," commonly called hallucinations, as well as "bias, and lack of transparency."

Parts of the alert follow, not the whole.

Consequences in the Real World

Courts, the alert says, "have increasingly sanctioned lawyers" for improper use of generative AI, mainly over fake citations and other hallucinations in legal filings, with typical penalties that have included financial sanctions, public censure, ethics courses and "disciplinary referrals to state bar authorities."

Those sanctions, in the alert's words, "serve as a warning for other tax professionals," and "Cases imposing sanctions against other types of tax professionals are slowly emerging." Its example is a report that a firm had prepared for the Australian government, which the government published in July 2025. The report "contained invented quotes attributable to a judge, references to non-existent reports, and books ascribed to the wrong author, all produced apparently by GAI." The firm, the alert says, reportedly resolved the matter directly with the government by agreeing to refund part of the fee it had received.

Circular 230, Read Against the Tool

Due diligence is Section 10.22. When using generative AI, practitioners "must thoroughly review all AI-created documents" before they reach a client or the IRS. "Due diligence requires verifying the accuracy of facts, citations, and calculations produced by AI." Practitioners "cannot rely solely on AI," and human scrutiny and editing are, in its word, "essential."

The alert's reading of Section 10.35, on competence, is that practitioners "must understand both the law and the technology used in their representation of clients before the IRS, including AI systems' operational mechanics, limitations, and risks," and be able to judge whether an output is fit for an IRS matter.

For written advice, the subject of Section 10.37, "practitioners cannot rely on GAI projections or representations without verification." "For legal documents, citations must be checked and cases read," it says. Where "the system's logic is opaque, reliance may be unreasonable under section 10.37." Practitioners should treat written advice drafted with generative AI "as a starting point, subject to thorough review before providing it to clients."

"GAI platforms may present risks regarding the unauthorized disclosure of sensitive taxpayer information, especially when data is uploaded to unsecured or public systems." Practitioners "must strictly handle all client data using only secure, enterprise-approved AI."

Every one of these duties lands on the practitioner. The tool can draft.

The Bill and the Firm

On fees, under Section 10.27, the alert says that because AI can cut research and drafting time, billing clients "for manual labor or time that was not actually spent or double billing for AI-assisted tasks may violate § 10.27, depending on the facts." Its examples of those facts are "a noticeable pattern across clients or the size of the billing differentials." Cost savings "should be passed on openly." Practitioners should disclose the AI activities performed, "in general or specific terms as needed," and fairly credit any cost reductions to the client's account.

That ties the speed of the tool to the bill. Where AI cut the hours, the bill is the record that shows whether the savings reached the client.

Under Section 10.36, those with "principal authority and responsibility for overseeing a firm's practice" must take reasonable steps to make sure the firm has adequate procedures. The alert adds that, for AI, firms must deploy policies whose coverage includes training for staff, protocols for "secure data handling, AI accuracy monitoring, etc.," and the vetting of outside tools. "And all steps and processes documented to show adherence to section 10.36."

Best Practices, and the Alert's Conclusion

The last two lines of its list of best practices tell practitioners to "Treat the written text that AI generates as drafts," and to review the resulting documents "for factual and legal accuracy (e.g., always check citations) and any problematic bias."

In its conclusion, "ethical obligations of competence, diligence, and confidentiality remain unchanged," and technology "serves as a powerful tool, not a substitute for professional judgment." It calls itself introductory, and notes that "directives and guidance from government agencies and professional bodies continue to evolve."

The last sentence reads, "Ultimately, responsible adoption of AI policies lies with the profession, to deliver enhanced value while preserving credibility."