ChatGPT and AI Tools in Your Tax Practice: What the IRS and Ethics Rules Actually Say
- Circular 230 §10.35 requires tax practitioners to exercise competence — that duty extends to any tool you use, including AI.
- IRS Publication 4557 outlines taxpayer data safeguard requirements that apply when using third-party software or cloud tools.
- A growing number of states have enacted comprehensive data privacy laws that can affect how you store and process client information — check the rules in your own state.
- A single unauthorized disclosure of a client’s SSN or financial data can trigger penalties under IRC §7216 — up to $1,000 per violation and/or one year imprisonment.
§1 What “Chatting AI” Actually Means for Tax Preparers
It’s a fair question. AI tools like ChatGPT, Google Gemini, Microsoft Copilot, and dozens of tax-specific AI assistants have become cheap, fast, and genuinely useful for drafting, research, and client communication. Millions of professionals use them daily.
But “using AI” covers a wide spectrum of risk. There’s a massive difference between asking an AI to “summarize the standard deduction rules for 2024” and pasting in a client’s name, SSN, income figures, and dependent information to get a return drafted. The first is research. The second may be a federal crime — or at minimum, a serious ethical violation.
As a tax preparer, you operate under a specific legal and professional framework. That framework doesn’t disappear because the tool is new and convenient. Let’s break down exactly what applies.
§2 The Circular 230 Competence Problem
Treasury Department Circular 230 governs practice before the IRS. Section 10.35 requires practitioners to possess the knowledge, skill, thoroughness, and preparation necessary for the matter at hand. Section 10.33 lays out best practices, including using reasonable care in all work product.
Here’s the trap: AI tools hallucinate. They confidently state incorrect tax law, cite nonexistent IRS rulings, and apply outdated figures. If you copy an AI’s answer into a client memo or use it as the basis for a return position without verifying it independently, you own that error. “The AI told me” is not a Circular 230 defense.
AI RISK MATRIX — COMPETENCE ISSUES BY USE CASE
| AI Use Case | Risk Level | Circular 230 Exposure |
|---|---|---|
| Drafting a client newsletter about general tax tips | Low | Minimal — verify figures before publishing |
| Researching whether a deduction exists | Medium | Always confirm against IRS.gov or primary source |
| Generating a return position or tax opinion | High | Significant — you sign the return, not the AI |
| Drafting IRS correspondence with client-specific facts | High | High — accuracy and privilege issues apply |
| Inputting raw client PII into a public AI tool | Severe | Potential §7216 violation — see §3 |
§3 IRC §7216 and the Data Privacy Red Line
This is the section that should make you pause before you hit “send” in any AI chat window.
IRC §7216 prohibits any person engaged in the business of preparing tax returns from knowingly or recklessly disclosing or using any information furnished to them for the purpose of preparing a return — for any purpose other than preparing that return — without client consent.
When you paste client data into a public AI tool, you are transmitting that data to a third-party server. Most consumer-facing AI products use your inputs to improve their models unless you specifically opt out or use an enterprise/privacy plan. That transmission may constitute a “disclosure” under §7216.
The IRS addressed third-party disclosure rules in Rev. Proc. 2013-14, which requires written client consent for most non-preparation uses of return information. A chatbot session almost certainly does not meet that standard.
Before your next AI session. Review the Terms of Service and privacy policy of any AI tool you use. Confirm whether it stores your inputs, trains on them, or shares them. If you cannot confirm data isolation, do not input client PII — period.
§4 Where AI Genuinely Helps vs. Where It Can Burn You
We’re not here to scare you away from useful technology. AI has real, legitimate applications in a tax practice — as long as you draw the right lines.
PRACTICAL GUIDE — AI GREEN LIGHT vs. RED LIGHT
| Task | Green / Red | Notes |
|---|---|---|
| Writing blog posts or social media content about tax topics | ✅ Green | Edit for accuracy; verify all figures |
| Summarizing a complex IRS notice in plain English for your own understanding | ✅ Green | No client data involved |
| Creating engagement letter templates or office policy documents | ✅ Green | Attorney review still recommended |
| Asking general “how does the QBI deduction work” questions | ✅ Green | Treat the answer as a starting point, not the finish line |
| Inputting a client’s name, SSN, income, or address | 🔴 Red | §7216 risk; use anonymized hypotheticals instead |
| Having AI draft a complete tax return or Schedule C | 🔴 Red | Accuracy unverifiable; you still sign under penalty of perjury |
| Relying on AI-cited “IRS rulings” without verification | 🔴 Red | AI frequently fabricates citations |
§5 A Practical AI Use Policy for Your Office
Whether you’re a solo preparer or managing a team, having a written AI use policy protects you, your clients, and your PTIN. Here’s a simple framework to adopt or adapt:
1. Anonymize before you inquire. Replace client names, SSNs, and identifying details with generic placeholders. “My client earned $82,000 in wages and has a home office” is fine. Pasting their W-2 is not.
2. Verify every figure and citation. Before using any AI-generated tax figure, threshold, or legal citation in client-facing work, confirm it against IRS.gov, the Internal Revenue Code, or a trusted tax research platform.
3. Choose privacy-forward tools. Enterprise versions of AI tools (Microsoft 365 Copilot for business, for example) often offer data isolation agreements. Evaluate these before using AI with any sensitive context — even anonymized.
4. Document your review process. If AI helped draft a memo or research a position, note in your workpapers what you verified independently. This demonstrates due diligence if a Circular 230 question ever arises.
5. Stay current. IRS guidance on AI and practitioner tools is evolving. Check IRS.gov and your professional association bulletins at least quarterly for updates.
Can I use ChatGPT to answer a client’s tax question?
You can use it to help you research or draft a response, but never feed in the client’s personal information, and always verify the AI’s answer against authoritative sources before passing it along. You are responsible for the accuracy of any advice you provide.
Does using AI for tax work require client consent?
If you are inputting client tax return information into a third-party AI tool, IRC §7216 and Treas. Reg. §301.7216 likely require written client consent. Consult the regulations and consider having your attorney review your process. When in doubt, anonymize the data first.
Are there AI tools built specifically for tax preparers that are safer?
Yes — several tax software companies are building AI features directly into their platforms with data agreements already in place. These are generally safer than consumer AI tools because the vendor has already addressed data handling in your existing service agreement. Review those agreements carefully and ask vendors specifically about data retention and training practices.
Will the IRS penalize me if an AI makes an error on a return I sign?
The IRS holds the preparer of record responsible for return accuracy, regardless of what tools were used. Accuracy-related penalties, preparer penalties under IRC §6694, and Circular 230 sanctions can all apply. “The AI got it wrong” is not a recognized defense.
Tax IQ Academy is an IRS and CTEC approved provider of continuing education for tax professionals, offering AFSP, AFTR, and EA exam prep courses covering federal tax law, ethics, and practice management.