Accounting Firms Face Data Exposure Risk From Unapproved AI Tools

Unapproved AI use can turn a busy season shortcut into a client data exposure. Accounting firms handle tax returns, Social Security numbers, bank details and business financials; when staff paste those records into a consumer chatbot, the firm may lose visibility into where the information goes and how it is handled.

The issue is already broader than individual chatbot use. “Shadow AI” includes personal accounts used to summarize returns, assistants connected to firm email or document storage without review, and apps given access to client folders without a partner’s approval. Each route can move sensitive information outside the firm’s established controls.

That creates an operational and reputational problem even when there’s no obvious breach. A firm may not know what employees uploaded, whether a service retained the material, or what terms applied to its use. If a client asks where their return went, the firm needs more than a general assurance that staff acted in good faith.

Why the exposure matters

Clients entrust accounting practices with records that can reveal identity, income, assets and business finances. Those details are valuable to attackers and data brokers, and clients expect firms to handle them carefully. The risk is not limited to an AI system producing an incorrect answer. Sending a complete return or bank statement to an unreviewed service creates uncertainty about access, retention and the firm’s ability to account for the data later.

A hypothetical busy-season example shows how easily the gap can form. A junior preparer uploads a client’s return PDF from a personal account to get help with a tax treatment question. The response is useful, so the preparer repeats the process for another client the next day. Nobody at the firm reviewed the tool’s data terms, and the firm has no record of what was uploaded or whether it was kept.

Nothing has to go visibly wrong for the practice to face difficult questions. The exposure may remain unnoticed until a client, insurer or other reviewer asks how the firm used the information. By then, staff may struggle to reconstruct which documents went into which services.

Firm duties and vendor oversight

Many accounting and tax firms are covered by the FTC Safeguards Rule, which calls for a written information security program, a risk assessment and oversight of service providers that handle customer information. An AI company processing client records may qualify as such a provider, so firms need to assess the service and its data practices as part of their security work.

Tax professionals can also consult the IRS’s Publication 4557, Safeguarding Taxpayer Data. Section 7216 limits how preparers may use and disclose tax return information. How those requirements apply to a particular firm depends on its circumstances; firms should confirm their obligations with professional advisors.

The practical point is that buying or enabling an AI service doesn’t settle the question of whether client information belongs there. Firms need to understand what the vendor does with submitted data, what controls administrators have, and whether the arrangement fits the firm’s security program and professional responsibilities.

Controls that close the gap

Start by classifying information. Tax returns, Social Security numbers, bank records and client financial statements should be off limits in tools the firm hasn’t approved. A written acceptable-use policy can make the boundary clear: name the kinds of information staff may not enter, identify permitted uses, and explain how employees can request review of a new tool.

Then choose approved services deliberately. Review them as you would other providers with access to client information. Business plans with administrator controls, multifactor authentication and activity logs can give the firm more oversight. Written terms should address data retention and whether submitted information may be used for training.

Take an inventory of existing AI use before the next filing deadline. Ask employees which tools they use and what work they perform with them. The quality of that inventory depends on staff feeling safe to answer honestly; a punitive tone may drive use further out of view. Inventory should include personal accounts, browser assistants and apps connected to firm email, shared storage or practice management systems.

Access controls matter as much as the written rules. Require administrator approval before an AI app can connect to email, document storage or practice software. DNS filtering can block or flag unapproved services on firm devices and networks, while activity logging can help identify unusual use that a policy alone won’t catch. Include AI access in staff training before busy season, and make sure offboarding removes departing employees from every tool tied to client data.

Keep records of the firm’s policy, approved tools, vendor reviews, staff training and monitoring. Those records help the firm explain its decisions if a client, insurer or other reviewer asks how it protects information. They also give leadership a way to spot gaps between the written rules and everyday practice.

Be ready to answer clients

Clients may ask whether their tax return is processed with AI, which tools the firm allows, and how the firm prevents employees from using personal accounts. Prepare direct answers that describe the approved services and the protections around client data. If consent or disclosure questions arise, confirm the applicable requirements with professional advisors and state the firm’s position in its engagement letter.

Clients don’t need their firm to avoid AI altogether. They need a clear account of where the technology is used and how their information stays protected. A short policy, a limited list of reviewed tools and basic monitoring give firms a practical basis for that answer—and make it easier to see where staff practices need attention.