For accounting firms
AICPA Confidentiality Rules and AI Tools
How the AICPA confidentiality rule, its third-party service provider interpretations and the due care standards apply when your firm uses AI tools.
Updated 6 min read
AI tools can draft memos, summarize documents and answer research questions in seconds. Each time someone at your firm pastes in client data, though, that data goes to an outside company. The AICPA Code of Professional Conduct already has rules for that situation, and this guide walks through them.
What the confidentiality rule says
The Confidential Client Information Rule (ET 1.700.001) is short. A member in public practice must not disclose confidential client information without the client’s specific consent. The rule has exceptions for things like professional standards, subpoenas, peer review and ethics investigations. None of them covers convenience or new software.
The Code defines confidential client information broadly (ET 0.400.11). It means any information obtained from the client that is not available to the public. The same definition reminds members that federal, state or local rules on confidentiality may be stricter than the Code.
When staff type a client’s numbers into a chatbot or upload a client’s documents, the AI provider receives that information. So the question is not whether a disclosure happens. It is whether you have met the conditions the Code sets for disclosing to an outside provider.
The three third-party service provider interpretations
The Code treats an AI provider the way it treats any outside party that helps you deliver professional services. Three interpretations work together, and each points to the other two.
Disclosing information to a third-party service provider (ET 1.700.040)
This interpretation sits under the confidentiality rule. It starts from the idea that clients may not expect you to use an outside provider. Before you disclose confidential client information to one, you should do one of two things:
- Sign a contract. Enter into an agreement with the provider to keep the information confidential. You also need reasonable assurance that the provider has procedures to stop unauthorized release of the information. How much checking that takes depends on the facts, including how much the provider publishes about its controls.
- Get specific consent. Obtain the client’s specific consent before you disclose.
For most firms, the contract route is the practical one. It means reading the provider’s business terms, not just its marketing page, and keeping a record of what you checked.
Use of a third-party service provider (ET 1.150.040)
This interpretation sits under the Integrity and Objectivity Rule. Before you disclose confidential client information to an outside provider, you should tell the client, preferably in writing, that you may use one. If the client objects, you should either not use the provider for that work or decline the engagement.
There is an exception. You do not need to tell clients about providers that give you administrative support, such as record storage, software application hosting or authorized e-file transmission. An AI tool that drafts analysis or reviews client documents is arguably helping you perform professional services, not just hosting software. The cautious approach is to give clients notice, for example in your engagement letter. The Code points members to nonauthoritative sample client disclosure language that the AICPA has published for this purpose.
Use of a third-party service provider (ET 1.300.040)
This interpretation sits under the General Standards Rule. It covers providers that help with bookkeeping, tax preparation, consulting or attest services. Before using one, you should make sure it has the qualifications, technical skills and resources for the work. You must also plan and supervise its work, get enough relevant data to support the work product, and follow the technical standards that apply.
Due care and the General Standards Rule
The Due Care principle (ET 0.300.060) asks members to follow the profession’s technical and ethical standards and to carry out their work with competence and diligence. Competence includes knowing your limits.
The General Standards Rule (ET 1.300.001) turns that principle into four requirements:
- Professional competence. Take on only work you can reasonably expect to complete competently.
- Due professional care. Exercise due care in performing the work.
- Planning and supervision. Adequately plan and supervise the work.
- Sufficient relevant data. Have a reasonable basis for your conclusions and recommendations.
AI tools can produce confident answers that are wrong, cite rules that do not exist or miss facts. Under these standards, AI output is a draft. A professional must check the facts, the math and the authorities before anything reaches a client or a return. The Code also links to a nonauthoritative Ethics Q&A on using the output of technology (section 400.02) and a staff article on AI through an ethics lens. They are worth reading, but they do not change the rules.
Practical steps for using AI under the Code
- List the AI tools in use. Ask staff what they use, including free chatbots, browser add-ons and AI features inside software you already pay for.
- Read the terms for the account you actually use. Personal and free accounts often have different terms from business agreements. Look for confidentiality commitments, whether the provider trains on your data, how long it keeps data and where it processes data.
- Put a written agreement in place. This supports the contract option in ET 1.700.040. Keep a short file with the agreement and the security information you reviewed, so you can show reasonable assurance later.
- Tell clients. Add a plain sentence to engagement letters saying you may use third-party service providers, including AI tools, and explain how a client can object.
- Share only what the task needs. Remove names, account numbers and identifiers when they add nothing to the question.
- Set a firm policy. Say which tools are approved, what data may go into them, who may use them and who reviews output. Train staff on it.
- Review and document. Have a professional review every AI-assisted work product, and keep the reviewed version in your workpapers.
- Check the other rules. Your state board of accountancy may have its own confidentiality rules. If you prepare tax returns, IRC §7216 and the FTC Safeguards Rule also apply, and they are stricter in places.
How Zrico fits
Zrico gives each firm its own private workspace with its own application and databases. Claude is the default model and is processed only in the US; GPT and Gemini are off unless the firm owner turns them on. No AI provider trains on your prompts, files or answers. Sign-in is invite-only with two-factor authentication, required for admins, and the owner gets a usage and audit report and retention settings. Zrico has no SOC 2, ISO or HIPAA certification and does not make a firm compliant, so do your own review under ET 1.700.040 and have a professional review all output.
Sources
- AICPA Code of Professional Conduct, online edition (ET 0.300.060, 0.400.11, 1.150.040, 1.300.001, 1.300.040, 1.700.001, 1.700.040): https://pub.aicpa.org/codeofconduct/Ethics.aspx
- AICPA Code of Professional Conduct, full PDF, updated through September 2026: https://pub.aicpa.org/codeofconduct/ethicsresources/et-cod.pdf
- IRS Office of Professional Responsibility, Issue Number 2026-19, Introductory Guidelines for Responsible AI Use in Federal Tax Practice (June 24, 2026): https://www.irs.gov/pub/opr-taxpros/issue-number-2026-19-introductory-guidelines-for-responsible-ai-use-in-federal-tax-practice.pdf
- 26 U.S.C. §7216, Disclosure or use of information by preparers of returns: https://www.law.cornell.edu/uscode/text/26/7216
- 16 CFR §314.4, Elements of the Safeguards Rule: https://www.law.cornell.edu/cfr/text/16/314.4
General information, not tax or legal advice. Check the current rules and your own facts.