Shadow AI in Professional Services Firms: The 2026 Risk Report
Published November 18, 2026 · Published November 2026 · By The Crossing Report · 9 min read
Eight or nine of your staff members are using AI tools you didn't choose, on client data you haven't secured, to produce work you haven't formally approved.
That's not a forecast. That's what the data says about the statistical average professional services firm right now.
Intapp's 2026 Technology Perceptions Survey — 1,800 professionals across accounting, law, and finance — found that 83% of accounting and finance professionals and 64% of legal professionals use AI tools their firm hasn't approved. The term for this is shadow AI: the AI your staff is running that you don't see, didn't evaluate, and may not know exists.
This page covers what shadow AI actually looks like inside a professional services firm, what specific laws and liabilities it triggers, why a policy alone doesn't solve it, and the five steps to close the exposure gap — most of which can be done in a single week.
What Shadow AI Looks Like in Practice
Shadow AI is not a dramatic event. It doesn't start with someone deciding to break a rule. It starts with a deadline and a tool that makes the work easier.
A paralegal is drafting a client summary with a document attached. The firm's approved AI tool is slow or limited on this task. They open ChatGPT — the free version on their personal account — paste the relevant section, and get a usable summary in 30 seconds. The work gets done. No one finds out.
A staff accountant is reviewing a complex client situation. They don't have a firm-approved research AI. They type the scenario into Gemini, get a reasonable answer, and move on.
An associate at a law firm is under pressure to get a contract memo out. They use a consumer AI tool to draft the first version — the same tool they use personally for writing help — and edit it into something they're satisfied with.
In each case: the person was not trying to cause a problem. They were trying to do good work efficiently with the tools they had. The problem is what happens to the client data they just submitted to an external service with no data processing agreement, no confidentiality protections, and potentially a training data use clause in the terms of service.
Why the Rate Is 83%
The 83% shadow AI rate in accounting and finance is not a failure of individual judgment. It's the structural result of a technology transition that happened faster than governance could keep up.
Wolters Kluwer's data: accounting firm AI adoption jumped from 9% in 2024 to 41% in 2025. That is a 32-percentage-point increase in a single year — the steepest one-year growth curve of any professional services vertical.
When adoption grows that fast, it means most of the current AI use at accounting firms happened before any governance framework existed. Staff found tools that helped. The tools were there. The policy wasn't. The staff policy gap driving shadow AI adoption is structural, not behavioral — and it won't close without a parallel investment in approved alternatives.
Add to this: Thomson Reuters' 2026 Future of Professionals Report found that 41% of professionals lack access to professional-grade AI tools that protect confidential data, use verified content, and produce explainable outputs. For those professionals — nearly half the field — shadow AI isn't rogue behavior. It's the only practical AI available.
The shadow AI problem in professional services has two causes. One is staff behavior. The other is a tool supply gap: firms that invested in governance expectations without investing in workable approved tools are partly responsible for the problem they're trying to solve.
The Federal Exposure Accounting Firms Should Know
For accounting firms, shadow AI with client tax data is not just a vague data security concern. It is potential criminal exposure under three specific federal provisions.
Section 7216 of the Internal Revenue Code prohibits unauthorized disclosure or use of tax return information. When a staff member submits client tax data to a consumer AI service — even without malicious intent — that is a potential Section 7216 violation. The penalties include fines and up to one year of imprisonment. The critical word is "unauthorized": if the AI tool is not an authorized, governed service with appropriate data protections, the disclosure is unauthorized. For a detailed breakdown of data exposure scenarios from unauthorized AI use at professional services firms, including specific disclosure patterns and what triggers each provision, see the accompanying analysis.
GLBA (Gramm-Leach-Bliley Act) requires financial institutions — including tax preparers — to maintain a written information security program and to protect nonpublic personal information. Consumer AI tools typically do not meet GLBA safeguard standards. Submitting client financial data to such tools is not compliant with GLBA's data handling requirements.
IRS Publication 4557, the IRS's guidance on safeguarding taxpayer data, requires firms to identify all third-party services that handle taxpayer data and ensure those services maintain appropriate security controls. Consumer AI tools are third-party services. If they're in use without being formally evaluated and documented in your data security plan, your firm is out of compliance with IRS 4557 guidance.
These are not theoretical risks. They apply to firms using unsanctioned AI on client data right now.
What ABA Opinion 512 Means for Law Firms
For law firms, the shadow AI exposure runs through professional responsibility rather than federal criminal provisions — but the practical stakes are equivalent.
ABA Formal Opinion 512, issued in 2024, applied existing Model Rules to AI tool use. The competency requirement (Rule 1.6) now explicitly includes understanding the AI tools your firm uses. If your staff is using AI tools you haven't evaluated, supervised, or governed, the firm is potentially failing the competency standard for every piece of work product that passed through that tool.
The confidentiality requirement (Rule 1.6) is the most direct shadow AI concern. Pasting client information into a consumer AI service with undefined data retention policies is a potential Rule 1.6 violation. The rule requires reasonable measures to prevent unauthorized disclosure. Consumer AI tools are not reasonable measures.
The supervision requirement (Rule 5.3) requires lawyers to make reasonable efforts to ensure non-lawyer assistants comply with professional obligations. AI tools your staff uses are, in the ABA's framework, tools that require supervision equivalent to the supervision of the work itself. You cannot supervise tools you don't know about.
The Insurance Risk at Your Next Renewal
Professional liability insurers started asking about AI governance at renewals in 2026. The question is not theoretical: they are asking for written AI policies and evidence that staff acknowledged them.
IBM's 2025 Cost of a Data Breach Report provides the financial backdrop. Organizations with no AI governance policy were breached at higher rates, and when they were, shadow AI added approximately $670,000 to the average breach cost. The mechanism: the breach scope was broader because the firm didn't know which client data had already been processed through unsecured third-party tools.
For a small professional services firm, the renewal conversation is becoming concrete. Insurers are categorizing firms with documented AI governance frameworks differently from firms without them. A written policy with staff acknowledgment is the minimum documentation threshold most are now looking for. Firms that cannot demonstrate that framework are seeing premium increases — and in some cases, coverage questions if a claim involves an AI tool.
The Five-Step Shadow AI Governance Fix
This is not a six-month project. For a 5-20 person firm, it's a week of focused work.
Step 1: Survey your staff without penalty. Ask which AI tools your team currently uses for work. Make clear this is not an audit — it's an inventory. You need accurate information more than you need compliance theater. Set a 48-hour deadline. The list of tools will almost certainly surprise you.
Step 2: Categorize by data risk. For each tool identified, determine whether client data flows through it. Consumer AI accounts (personal ChatGPT, free Gemini, free Claude) are the high-risk category — no data processing agreement, potential training data use, no confidentiality protections. Enterprise tools with data processing agreements — Microsoft Copilot under M365, CoCounsel under Westlaw, Checkpoint AI features — are a different risk category. You need to know which bucket each tool falls into.
Step 3: Write a one-page AI policy. State which tools are approved for work use. State explicitly that client data may not be processed through non-approved tools. State what review is required for AI-assisted work product before it goes to a client. The policy does not need to be long — it needs to be specific, findable, and have a date on it. For accounting firms starting from scratch, the CPA firm AI policy template provides a ready-to-adapt structure that covers OPR and Section 7216 requirements in one page.
Step 4: Get documented staff acknowledgment. An email to each staff member attaching the policy, with a reply confirming receipt and understanding, is sufficient. This is the documentation insurers are asking for. It takes 15 minutes to send and creates a record that your firm took the governance step.
Step 5: Provide a sanctioned alternative. If the survey revealed that staff are using shadow AI because they don't have an approved tool that handles their actual use cases, the policy alone will produce nominal compliance at best. Identify the most common use case driving shadow AI — usually drafting, research, or summarization — and find one approved enterprise AI tool that handles it with appropriate data protections. This is the investment step, but it is what makes the governance real rather than aspirational.
What Comes Next
The shadow AI problem is not going away on its own. Thomson Reuters estimated that $143 billion in US professional services client revenue is at risk for firms that fail to implement AI effectively — not from AI disruption, but from AI governance failure. The clients who eventually discover their data was processed through unsecured AI tools are clients who leave.
The good news: the governance fix for a small firm is genuinely achievable in a week. The barrier is not complexity — it's that most firm owners haven't made it a priority yet.
The five steps above are that priority. Survey, categorize, write, acknowledge, provide. One week of focused work closes the exposure gap that 83% of accounting firms and 64% of law firms currently have.
The formal guidance — from OPR, the IRS, state bars, and PI insurers — is tightening. Firms that act now are ahead of the requirements. Firms that wait will be reacting to them.
This is the kind of intelligence premium subscribers get every week.
Deep analysis, cross-sector patterns, and the frameworks that help professional services firms make the crossing.
Related Reading
- Your Firm Is in the 83%. Here's the Governance Framework to Get Into the 25%.
- Your Cyber Insurance Renewal Is About to Ask About AI — Here's What to Prepare
- Do Professional Services Firms Need to Disclose AI Use? (2026)
- AI Use Policy Template for Small Law Firms (2026)
- Accounting Firm AI Policy Template (2026 Edition — Copy and Customize)
- Microsoft Copilot Gets Accounting Skills This Week — What CPA Firms Need to Do
- The IRS Just Made AI Billing an Ethics Issue — What Every CPA Firm Needs to Do This Month
- The IRS Has No Rules for AI Tax Prep — And That's Your Firm's Problem, Not Theirs
- Law Firm AI Governance Has a Documentation Problem — AllRize's GRC Module on Microsoft Purview Is One Answer
- The Official Internal Controls Framework for AI Is Here — What COSO's New Guidance Means for CPA Firm Owners
- Google Workspace vs Microsoft 365 for Professional Services Firms in 2026: The AI Cost Breakdown