Accounting and tax

AI for accounting and tax firms, with client data kept out.

Updated September 24, 2026 · SOPHIA XT LLC

The risk

Staff draft client emails, engagement letters and notice responses with AI because it saves time. The fastest way to do that is to paste the client's details in. Once a return figure or an SSN is in a prompt, it has been disclosed to the AI provider, whatever that provider later does with it. If the staff member used a personal account, the firm has no record that it happened.

The rules that apply

IRC 7216. A tax return preparer who knowingly or recklessly discloses or uses tax return information for anything other than preparing the return commits a misdemeanor, punishable by a fine of up to $1,000, up to a year in prison, or both (26 U.S.C. 7216). The civil side, 26 U.S.C. 6713, is $250 per disclosure, up to $10,000 a year. The regulations at 26 CFR 301.7216 define tax return information broadly and set out when written client consent allows a disclosure. The IRS keeps a Section 7216 information center.

The FTC Safeguards Rule. Tax preparers and many accounting firms are financial institutions under the Gramm-Leach-Bliley Act, so the Safeguards Rule (16 CFR Part 314) applies. It requires a written information security program, oversight of service providers that handle customer information (section 314.4(f)), and, since May 13, 2024, notice to the FTC within 30 days of discovering an event that exposes unencrypted information of 500 or more consumers (section 314.4(j)). The FTC's plain-English summary is here.

IRS guidance. Publication 4557 (Safeguarding Taxpayer Data) and Publication 5708 (a sample written information security plan) are the IRS's own checklists for small practices.

None of these rules names ChatGPT. Each asks you to control who sees client tax information and to keep a record. An AI provider that receives placeholders instead of values has not received the tax information.

How Safe-Desk works in a tax office

Staff keep using ChatGPT, Claude, Gemini, Copilot, Perplexity or DeepSeek. The Safe-Desk extension for Chrome and Edge, or the Safe-Desk Windows app for the ChatGPT, Claude and Copilot desktop apps, checks each message before it leaves. Client data becomes a placeholder such as [SSN_1], and only that version reaches the AI provider. When the answer comes back, the staff member who sent it sees the real values again; nobody else does. Passwords, keys and text marked privileged are stopped and sent nowhere. The owner gets a log of who sent what kind of data and which rule caught it, never the value itself, plus phone alerts and a monthly report. See how it works and the security page.

What the accounting template catches

Pick Accounting and tax at sign-up. These are the defaults; the owner can change names, places and dates to off, flag or protect, and add a client list and custom topics.

Read from the Accounting and tax preparation template in the product. In strict mode, messages with identifiers such as SSNs and account numbers are stopped instead of swapped.

A worked example

The staff member reads the answer with the real values back in. The log names what kind of data was caught, never the values.

What it does not do

Safe-Desk does not make a firm compliant on its own and does not certify compliance. It has not been audited for SOC 2. It does not sign a Business Associate Agreement, so it is not for patient information. Names, addresses and dates are caught from your client list and from context; context detection is a best effort, because names have no checksum. Anything your firm sets to Flag is sent as written and reported to the owner.

Safe-Desk does not replace the written consent 7216 requires for disclosures you choose to make, and it does not replace your written information security plan. It gives that plan a control for AI use and a record you can show.

Setup and price

An owner can sign up and pick the template in a few minutes. The extension goes on each computer in about five minutes today, from a download in Settings, because it is not in the Chrome Web Store yet. Safe-Desk costs $19 per staff member a month, with a 14-day free trial. Each seat includes $5 of AI a month inside Safe-Desk, pooled across the firm, and AI beyond that is billed at the model's cost plus 30%. Using your own ChatGPT or Claude through the extension or the connector costs nothing extra. Optional setup: we set it up for $149 for up to 10 staff, $10 for each extra person.

Related

• Is it safe to use ChatGPT with client data?What happens to what you type, business versus consumer plans, and the rules that apply.
• A ChatGPT and AI use policy for small firmsWhat the policy should say, and a one-page sample to copy.
• How Safe-Desk compares with the other ways to keep client data out of AIEnterprise DLP, browser-only extensions, private workspaces and banning AI, side by side.
• SecurityWhat we keep, what we never keep, and every subprocessor.