Using AI for tax advice is not automatically malpractice. The risk comes from relying on its work without adequate review, mishandling client information, or presenting advice that does not meet the professional standards already in place.
On June 24, 2026, the IRS Office of Professional Responsibility addressed that point directly in guidance for federal tax practice.
What the IRS guidance says
The Office of Professional Responsibility, which investigates Circular 230 violations, issued Alert 2026-19, “Introductory Guidelines for Responsible AI Use in Federal Tax Practice.”
The alert warns about fabricated output, bias, and lack of transparency. It also tells practitioners to review documents created with generative AI carefully.
The tool may draft the work, but the practitioner remains responsible. The alert maps that responsibility to existing rules rather than creating a separate AI standard.
How Circular 230 applies
Each of these duties predates ChatGPT. AI changes the workflow, not the standard.
| Circular 230 | Duty | What it means when AI drafts the work |
|---|---|---|
| § 10.22 | Diligence as to accuracy | Verify every fact, citation, and calculation the model produced before it reaches a client or the IRS. |
| § 10.35 | Competence | Understand the tool’s mechanics, limits, and failure modes — not only the law. |
| § 10.37 | Requirements for written advice | Base advice on assumptions and citations you have checked, not on the model’s confidence. |
| § 10.36 | Procedures to ensure compliance | Partners must set firm policy: staff training, approved tools, data-handling protocols, documentation. |
| § 10.27 | Fees | Do not bill for manual labor or time that was not actually spent, or double bill for AI-assisted work; the alert says whether a fee violates the rule depends on the facts. |
| § 10.51(a)(15) | Confidentiality | Use appropriately secured, firm-approved systems and follow the rules governing disclosure or use of tax return information. |
Confidentiality also extends beyond Circular 230. Improper disclosure or use of return information may implicate IRC §7216 and the civil penalty under §6713. Firms should approve tools and data-handling practices before client information enters the workflow.
So is it malpractice?
Short version: the OPR alert does not say that using AI is itself malpractice. It says the practitioner’s existing duties of diligence, competence, confidentiality, and care in written advice continue to apply. Whether conduct amounts to malpractice depends on the facts and applicable law.
The practical line is professional sign-off. When advice goes to a client or a filing reaches the IRS, the practitioner is responsible for the facts, assumptions, citations, and calculations behind it. AI cannot exercise that diligence on the practitioner’s behalf.
Why verification can take longer than expected
Checking a general model’s answer may mean rebuilding its research. A model can generate a plausible Code section or invent a Tax Court case without having retrieved either source. A reviewer then has to determine whether the citation exists before asking whether it supports the claim.
That is why verification should run against current primary authority, not the model’s confidence or the reviewer’s memory.
Give the model a source you can open
Retrieval makes that review easier. When the tool returns a citation from an index of real authority, the reviewer can open the source, read the relevant text, and test the analysis. Retrieval does not guarantee that the research is complete or correctly applied, but it removes much of the work involved in reconstructing a bare citation.
That is what TaxMCP is built to do. Results include a link to the IRC section, Treasury regulation, IRS publication, ruling, or Tax Court opinion used in the response.
TaxMCP does not require a client return or supporting documents, and it does not persist MCP tool arguments in its application database. Search text is processed through OpenAI’s API for embeddings and, when enabled, reranking under OpenAI’s API data controls. That can help a firm minimize the information sent to an additional research service, but it does not make the surrounding Claude or ChatGPT workflow compliant by itself. Information a practitioner enters into the AI client remains governed by that provider’s terms, the firm’s policies, and the practitioner’s applicable obligations. See the full security and data-handling description.
The source link does not replace professional review. It gives that review a concrete starting point.
The argument worth having now
The alert leaves responsibility with the practitioner. The practical challenge is making primary-source review routine enough to perform on every AI-assisted engagement.
See how taxmcp grounds every answer in primary authority →
Or start with the broader question the alert forces: can you actually trust AI for tax research?