Artificial intelligence (AI) is quickly becoming part of everyday tax practice, but the IRS is making clear that using AI is not a substitute for professional judgment.gettyArtificial intelligence (AI) is quickly becoming part of everyday tax practice. It can help sort documents, speed up research, draft client letters, summarize facts, and make routine tasks easier. But the IRS is making one thing clear: using AI does not change a tax professional's responsibilities.In new guidance focused on responsible AI use in federal tax practice, the IRS Office of Professional Responsibility (OPR) reminds practitioners that the old rules still apply. AI may be new, but it does not replace professional judgment, and it does not excuse mistakes. For tax pros, that means Circular 230 still matters.What Is Circular 230?Circular 230, formally titled Regulations Governing Practice before the Internal Revenue Service, is the set of rules that governs practice before the IRS. It applies to tax pros who practice before the IRS—generally attorneys, certified public accountants, enrolled agents, and certain other authorized practitioners. Practice before the IRS can include communicating with the agency on behalf of a client, preparing and filing documents, giving written advice, and representing a taxpayer in IRS matters.It’s not just a “best practices” document. It’s actually codified in the federal regulations as Title 31, Subtitle A, Part 10. The rules have been around in some form for a long time, with the underlying statute dating to 1884 and Treasury first packaging the regulations as a Circular in 1921. Since then, the rules have been amended several times, including significant revisions in 2005 and 2014. The 2005 revisions are remembered largely for the covered-opinion rules and the wave of Circular 230 disclaimers that followed (though, the IRS has since implored you to ditch them—really). Overall, for tax pros who represent clients before the IRS, it sets the professional ground rules.Tax pros are expected to be competent, careful, honest, and responsive. They must exercise due diligence when preparing tax returns and other documents submitted to the IRS, be careful when giving written tax advice, provide records when properly requested, and avoid misleading the IRS or their clients. They cannot take frivolous positions, mishandle client funds, or represent clients when a conflict of interest prevents them from doing the job properly. Those rules may sound basic, but they are the framework that allows taxpayers, practitioners, and the IRS to rely on the integrity of the system.AI Can Help, But It Can Also Go WrongWhen it comes to AI, the concern is not that tax professionals are adopting new technology (that’s a good thing), but that a tax pro might treat AI as though it can take over the professional obligations that belong to the person using it.And AI is already here. Today’s legal research platforms, document review tools, workflow systems, tax software, and drafting tools all may include some form. Generative AI (GAI), which creates original content by relying on patterns, has accelerated use because these tools can create text, draft letters, summarize documents, answer questions, and suggest analysis. That can make tax pros more efficient, but it cannot be a substitute for exercising professional judgment. Under Circular 230, that judgment remains the practitioner's responsibility.You Still Have To Check The WorkCircular 230 requires tax pros to perform due diligence, meaning they must take reasonable care when preparing returns, documents, and other papers related to IRS matters, and must verify the accuracy of statements made to clients and to Treasury.Circular 230 also has specific rules for written advice on federal tax matters, including advice delivered by email. Those rules require practitioners to use reasonable factual and legal assumptions, consider the relevant facts, make reasonable efforts to identify those facts, and connect the law to them. Those obligations don’t change just because AI is involved. A tax pro who asks an AI tool to summarize a tax issue or calculate a potential tax liability cannot simply assume the answer is right. They must check facts and citations and (importantly), read cases, code sections and regulations. If a conclusion appears in a client memo or IRS submission, the tax pro should be able to explain how that conclusion was reached.Competence Includes Understanding The ToolCircular 230 has always required practitioners to be competent which means they must have the right level of tax knowledge, skill, preparation, and judgment. When it comes to AI, competence also means understanding the tools being used.Courts have sanctioned lawyers for improper AI use, most often because filings included fabricated cases, invented quotes, or other AI-generated errors. Those sanctions have included steep monetary penalties (sometimes several thousand dollars), public criticism, required ethics training or professional responsibility courses, removal from representation, and disciplinary referrals to state bar authorities.Those cases are not all tax cases, but tax pros should pay attention to them. And potential and alleged AI misuse is being flagged in the tax world. For example, OPR points to a July 2025 matter involving Deloitte Australia, which had prepared a 230-plus page report for the Australian government. The report contained invented quotes, references to nonexistent reports, and books ascribed to the wrong author, all apparently produced by GAI. Deloitte Australia reportedly resolved the matter by agreeing to partially refund a portion of the fee it received from the government.How you access and use AI also matters—a closed, secure research tool is different from a free public chatbot. That was part of the concern in a securities fraud case, in which a New York federal judge ruled that a criminal defendant’s conversations with the AI platform Claude about potential defense strategy were not protected by attorney-client privilege and did not qualify as protected work product. Although the ruling came in a criminal case, the same reasoning could have implications in civil matters as well—including disputes involving taxes. Asking an AI tool legal questions is not the same as speaking with counsel, and AI chats may be discoverable or used against the person who created them.This also ties in to client information. Tax pros routinely handle sensitive data like Social Security numbers and taxpayer business records, and federal law limits how that information may be used or disclosed. Uploading a client’s tax records, transcript, or return information into an unsecured public tool may create privacy and confidentiality problems. The safer approach is to use secure, enterprise-approved tools with clear data protections, and to understand whether the vendor uses that data for training, who can access it, and what happens in the event of a breach.Firms Need AI PoliciesThe IRS guidance also addresses the need for firm-level policies. Circular 230 requires firms to take reasonable steps to ensure there are adequate compliance procedures in place. A good AI policy should note which AI tools are approved, what kinds of client information may be entered, and what review is required before AI-generated work goes to a client or the IRS. It should also consider who is responsible for checking citations, how AI use should be documented, what training staff should receive, and how vendors are vetted. Firms also need to pay attention to emerging rules. Several states, including California, Colorado, Illinois, and Utah, have enacted AI governance legislation focusing on transparency and protecting consumers. In addition, professional organizations including the American Bar Association (ABA) have released guidance on the growing use of GAI in legal practice.Billing Needs Attention, TooAI may also affect billing. Circular 230 has rules that prohibit “unconscionable fees” in matters before the IRS. Since AI can reduce the time required for some tasks, the guidance suggests that firms should be thoughtful about how they bill for it. The issue is not that AI-assisted work has no value because it’s clear that professional judgment, review, strategy, and experience all retain value. But billing clients for hours that were not spent, or charging fees as though a task was completed manually when it was largely automated, is problematic. Billing should be transparent and understandable, no matter whether it’s for human hours or AI-generated content. At a minimum, firms should be able to explain what they did, what the client was charged for, and how the bill was calculated.The Bottom LineIt’s clear that AI can be a valuable part of tax practice. But it does not change the basic obligations of Circular 230. Tax pros still have to be competent, exercise due diligence, protect client information, give careful written advice, charge appropriately, and maintain adequate firm policies. Opting into tech use doesn’t mean opting out of human judgment.ForbesHow Confident Can You Be In The Person Preparing Your Tax Return?By Kelly Phillips ErbForbesTaxpayers Are Asking AI For Help. Trusting It Is Another Story.By Kelly Phillips Erb