Your staff want AI for client emails, research and reconciliations. AILeakShield shows you which AI tools are in use and stops Social Security numbers, bank details and client records before they reach the ones that shouldn't have them.
Sample data shown. Select the image to enlarge it.
It usually happens when someone is trying to get through the day's work faster.
A W-2, 1099 or organizer is uploaded so a tool can pull out the numbers.
A message with account details is pasted in to draft a reply.
A general ledger or bank statement is attached for reconciliation or analysis.
Employee names, Social Security numbers and pay are shared for a quick check.
Someone uses a personal AI login on a firm computer, outside anything the firm has reviewed.
A note taker or AI assistant connected to Microsoft 365 can reach mail, calendars and files.
The IRS says federal law requires tax and accounting professionals to create and maintain a Written Information Security Plan, because the Gramm-Leach-Bliley Act treats them as financial institutions. The FTC Safeguards Rule asks covered firms to keep a periodic inventory of data, monitor when authorized users access customer information, and oversee service providers, with some provisions exempted for firms holding information on fewer than 5,000 consumers. Separately, 26 U.S.C. 7216 limits how return preparers may disclose or use tax return information. None of these name AI tools, so ask your adviser how they apply to each tool your firm uses.
A technical control for a channel that many written plans don't cover yet.
Blocks financial data, personal identifiers and credentials, and warns on health data. You see every change before it's applied.
Detects Social Security numbers, ITINs, EINs, and bank account and routing numbers.
Inspects attachments, and reads the text in screenshots and scanned PDFs.
A risk-ranked list of the AI in use, ready for your risk assessment and data inventory.
See whether each vendor trains on input, where it's based and its SOC 2 status.
A record of decisions and blocked sends that doesn't contain the client's data.
Nothing changes for staff until you decide.
Start with the free Microsoft 365 report, with nothing to install.
Deploy to managed browsers in monitor mode. Nothing is blocked.
Approve the tools you trust, restrict the rest, and turn on your data rules.
Decisions, overrides and sign-offs are recorded for your next review.
No. Your plan is your own document, and IRS Publication 5708 offers a template for smaller practices. AILeakShield gives you a control and evidence to describe in that plan.
No. The written program, qualified individual, encryption and multi-factor authentication requirements are outside this product. AILeakShield covers one channel, which is AI tools in managed browsers.
No. Clearly labeled numbers are the strongest case. A number with no label around it may warn rather than block, and client details written in plain sentences can get through. It works alongside training and your AI policy.
Yes. Text going to an AI tool is checked by our service first, so include us in your vendor oversight. Blocked content is never stored, other prompt text is cleared at the end of each day (UTC), and our security page has the details.
AILeakShield doesn't require you to ban AI. Approve the web tools your firm has reviewed. AI features inside desktop tax software are outside what the browser protection can see.
Protection applies in managed browsers. Activity on unmanaged devices and personal phones isn't seen.
Start with a free report of the AI connected to your Microsoft 365. It doesn't read any email, files or chats.