New IRS AI Alert May Apply to Wealth Manager Practices
The Internal Revenue Service Office of Professional Responsibility has issued new guidance on the use of artificial intelligence in federal tax practice. The guidance, OPR Alert 2026-19, is directed to tax practitioners, but its practical impact may reach well beyond traditional tax preparers. Financial advisors, estate planners, attorneys, CPAs, enrolled agents and other professionals who use AI in connection with tax analysis, tax returns, tax projections, tax advice or communications with the IRS should pay close attention.
The reason is simple. Many advisors now use AI to review client documents, summarize tax returns, identify planning opportunities, generate client memoranda, draft explanations and organize financial information. Some advisors may upload tax returns, K-1s, estate tax information, trust income tax data, charitable planning documents or business records into AI platforms. Those activities may implicate tax return information, tax advice or matters administered by the IRS. If so, advisors may find themselves closer to the IRS professional-responsibility framework than they realized.
Risks of AI
The alert emphasizes that AI is already widely embedded in professional practices. Financial planning software commonly uses AI to analyze tax data and provide planning suggestions that advisors can give to their clients. AI tools can improve efficiency, accelerate data analysis and assist with drafting and other planning tasks. But the alert also identifies significant risks, including hallucinations (that is, fabricated authorities), bias, lack of transparency, confidentiality concerns and data protection issues. The alert’s basic message is to be careful and protect confidentiality; although AI can assist professional judgment, it can’t replace it.
The more difficult issue is how demanding the IRS standard may be.
Understanding of Operational Mechanics Required
The alert may create a tougher standard for tax professionals than the more flexible ABA approach, which generally asks whether a lawyer has kept reasonably current with technology, understands relevant benefits and risks, protects client information, supervises staff and vendors and uses professional judgment. The alert states that practitioners must understand both the law and the technology used in representing clients before the IRS, including AI systems’ operational mechanics, limitations and risks. It further states that practitioners must understand how AI develops content, recognize potential bias or errors and evaluate whether AI outputs are suitable for IRS matters. That language matters. “Operational mechanics” isn’t the same as a reasonable “understanding of benefits and risks.”
The alert doesn’t define how much technical knowledge is enough. Must a practitioner understand how a large language model is trained under the new alert? Must the advisor understand data retention policies, prompt creation, model drift, retrieval systems or vendor security architecture? Can vendor explanations, continuing education or consultation with technology staff suffice to meet the training requirements? The Alert doesn’t answer those and other important questions.
That uncertainty could be important for financial advisors. Consider a wealth manager who asks an AI platform to review a client’s tax return to identify Roth conversion opportunities, charitable planning issues, capital loss carryovers, estimated tax concerns or estate planning opportunities. If the advisor is merely using the tax return for investment planning, the advisor may view the task as financial planning. But if the work touches tax return information, tax advice or a matter potentially involving the IRS, the advisor shouldn’t assume the alert is irrelevant.
Due Diligence
The alert also addresses due diligence. Practitioners shouldn’t unquestioningly accept AI-generated facts, citations, calculations and conclusions. No responsible advisor should send a client an AI-generated tax analysis without review. The practical problem is how much review is sufficient. The alert states that AI work should be thoroughly reviewed and that factual and legal information should be verified. But it doesn’t clearly distinguish between a low-risk administrative use of AI and AI output that becomes written tax advice, a tax filing position or a communication to the IRS. If a financial planner has a limited tax background and uses common industry AI software to analyze a tax return and generate tax planning ideas, can that advisor thoroughly review the AI output? If not, are they in violation of the IRS requirements?
That distinction is important in wealth management. A planning team might use AI to summarize a 60-page tax return for internal review. It might use AI to draft a client-friendly explanation of capital gain exposure. It might use AI to identify possible issues for discussion with the client’s CPA. Those are different tasks with different risk levels. A sensible AI compliance policy should recognize those distinctions and ensure that protocols are established to ensure review of the AI output as to the facts used, the law applied, any AI citations to authorities relied on, and that security is maintained.
Confidentiality Standards
The alert’s confidentiality language is also significant. It warns against uploading sensitive taxpayer information into unsecured or public AI systems and states that client data should be handled only through secure, enterprise-approved AI with “robust” confidentiality safeguards. That is a prudent warning. Public AI platforms shouldn’t receive client tax returns or other sensitive client financial information, and even subscription-based AI shouldn’t receive confidential data unless the firm has carefully reviewed and approved the platform for that purpose, including review of the platform’s terms of service.
But again, the standard isn’t fully defined. What does “robust” mean? Does it require contractual restrictions on data training? Does it require enterprise licensing, encryption, access controls, audit logs, security certifications, data residency restrictions or all of these? ABA guidance tends to ask whether safeguards are reasonable under the circumstances. The Alert uses stronger language without providing the same balancing framework. So, practitioners should be alert to potentially stricter rules applying to AI tax matters than they may have anticipated.
Billing
The alert states that AI can reduce research and memorandum drafting time and that billing for manual labor or time not actually spent may raise concerns. It also says that cost savings should be passed on openly and that billing should reflect efficiencies gained through AI. No professional should bill fictitious time. If an hourly billing entry says six hours were spent manually reviewing documents when AI performed the initial review and the advisor spent one hour reviewing the output, that bill may be misleading. But many firms don’t bill for all services on a pure time basis. Some use fixed fees, planning fees, consulting fees or charges separate from asset management fees.
The alert creates questions for advisors who charge separately for AI-assisted tax planning. Suppose a financial advisor charges an additional planning fee to review a client’s tax returns with AI and identify potential income tax, estate tax or charitable planning opportunities. If AI reduces staff time, must the advisor reduce the fee? What if the advisor has invested substantial nonbillable time in selecting AI tools, reviewing terms of service, developing prompts, training staff, documenting procedures, securing client data and supervising output? The alert discusses cost savings, but it doesn’t define cost.
That omission is important. AI isn’t free merely because it reduces task-level time. Responsible AI use will require continuing education, vendor due diligence, cybersecurity review, staff training, prompt development, quality control, engagement-letter revisions and documentation. A narrow focus on reduced labor time may understate the real cost of responsible AI implementation.
Practical Steps
Advisors should consider several practical steps now:
-
Identify where AI is already being used in tax-related work.
-
Prohibit the use of public AI for tax returns and confidential client information. No AI should be used unless specifically approved under firm policy.
-
Review vendor terms, including confidentiality provisions, data retention and training provided.
-
Create a written AI policy that distinguishes low-risk internal uses from client-facing tax analysis.
-
Document human review of material AI-generated tax conclusions.
-
Revisit engagement and fee agreements and, in particular, disclosures to address AI use, tax-return work and billing.
-
Reconsider separate planning charges to ensure that they reflect real work, professional judgment, firm investment and client value, not artificial time entries.
Risk Management
The alert shouldn’t cause advisors to avoid AI. Used carefully, AI may improve client service and allow deeper analysis than clients might otherwise authorize. But AI can’t be treated as a casual shortcut. For advisors who work with tax returns, estate tax data, trust information, charitable planning, business ownership records or IRS communications, AI governance is now part of professional risk management.
Takeaway
AI shouldn’t be abandoned. Instead, advisors need a governed process: approved tools, trained personnel, protected data, reviewed output, clear billing and documented judgment. The IRS has entered the AI conversation. Wealth advisors who touch tax information should assume that conversation may include them.
