OPR Issues Guidance on Practitioner Use of AI

Every practitioner should know this guidance before using AI in their tax practice

The IRS Office of Professional Responsibility quietly sent an email to practitioners about the ethical issues of using artificial intelligence (AI) tools in tax practice.

Here’s the truth: OPR’s application of Circular 230 to AI tools makes absolute sense; however, my impression from talking to and seeing what practitioners claim is that they are not consistently following these guidelines with respect to client tax output.

Six Tax Provisions and Their Impact on AI

§10.22 – Due Diligence. A practitioner cannot rely on AI to do their job; they must review every AI-generated document before it reaches a client or the IRS, verifying facts, citations, and calculations. If the output looks plausible and verification is skipped, the practitioner has failed their due diligence obligations.

§10.27(a) – Fees. If an AI tool cuts research or drafting time in half but the practitioner bills as though they did the work without AI assistance, it can amount to an unconscionable fee. OPR expects the practitioner to disclose AI use and pass on cost savings to clients openly.

As AI tools make work much more efficient, it makes a compelling argument for switching from input-based billing methods, such as hourly billing, to fixed-fee value-based engagements in which the price is not dependent on practitioner inputs. This switch allows the practitioner to earn similar amounts with less work, enabling them to meet their §10.22 due diligence requirements better.

§10.35 – Competence. OPR has long insisted that technological competence is essential for tax practitioners, and now they are expanding this to understanding how AI works: its mechanics, limitations, and failure modes. If a practitioner cannot explain how a generative system produces content, recognize bias in its output, or judge whether a result is suitable for an IRS matter, the practitioner lacks the competence Circular 230 requires, and cannot deploy AI tools in their practice.

§10.36 – Firm Procedures. If a practitioner owns a firm, they are responsible for adequate procedures: staff training on AI risks, secure data-handling protocols, accuracy monitoring, vetting of third-party tools, and documentation showing that these requirements were met. “We didn’t know our staff was using ChatGPT for client advice” is not an excuse — have a standard operating procedure on AI use in place!

§10.37 – Written Advice. Every factual assertion and legal citation in AI-assisted written advice must be independently verified. You cannot rely on AI output mindlessly, especially when the underlying logic is opaque or when it involves a highly complex tax matter with detailed fact patterns. Treat any AI tax research tool as a starting point for the research process, not the end of it.

IRC §§6713 & 7216(a) – Data Protection. Civil and criminal penalties apply for unauthorized disclosure of tax return information. Uploading sensitive client data to unsecured or public AI platforms violates the law. Use only secure, enterprise-approved systems with real confidentiality safeguards. Or, better yet, avoid the issue altogether and do not upload sensitive tax return information into an AI tool.

Disclosure under §7216 is an essential issue, and OPR’s guidance is vague at best. More specific guidance is needed on how the §7216 exceptions that allow disclosure without consent apply to AI tools. See this previous article on this topic.

Responsible Adoption of AI in Tax Practice

AI is genuinely useful. It accelerates research, speeds drafting documents, and can surface relevant authorities. It can also help, or even take over, some firm administrative tasks. Firms that integrate it thoughtfully and ethically will outpace firms that do not.

The firms getting this right are:

• Selective about delegation. Not everything can be replaced with an AI tool; the human touch matters. Be extremely wary of AI tools touching sensitive client data, and a practitioner cannot outsource their tax judgment to an AI tool.

• Transparent with clients. Disclose that AI may be used in engagement letters. If §7216 protected information is disclosed to an AI tool, a practitioner must obtain specific, signed consent absent other IRS guidance.

• Disciplined about data disclosure. Sensitive information must stay in secure systems; public platforms are off-limits.

• Training and superving. Staff must understand the law, the ethical obligations, the technology, and the firm’s procedures, and supervisors spot-check usage.

• Documenting. Procedures, training, and verification are documented in standard operating procedures. In addition, the firm's written information security plan (WISP) is up to date regarding AI usage and the related data security safeguards.

This guidance tells practitioners that the use of AI in tax practice is on OPR’s radar. It is time to comply with best practices, even if it is not the most efficient use of AI tools.

A Circular 230 Compliant Tax Research Tool Exists

Josh Youngblood, EA, CRETS and I have been worried about the impact of AI on tax research — quick answers, typically unverified, that do not meet Circular 230 requirements or ensure taxpayers get a quality research process.

We built the first AI tax research tool — AskTomG.ai — with safeguards built in to ensure compliance with Circular 230 requirements and a thorough tax research process. It uses the same tax research process that I teach and have used for years.

AskTomG.ai finds the authority, but you make the call. It surfaces relevant authorities, tags them by weight under Treas. Reg. § 1.6662-4(d)(3)(iii), and hands you a vetted foundation for your own analysis. The judgment stays with you—where Circular 230 says it belongs, and OPR has confirmed.

AskTomG.ai was launched last week. Become a Founding Member by June 26, 2026, for a two-year price guarantee and special benefits!

Join the Conversation

If you are a paid subscriber, you can talk about this topic in the comments section. Please keep the discussion related to this edition’s topic.