SAMPLE WISP · 2026 · WITH THE CIRCULAR 230 AI SECTION
A finished Written Information Security Plan, all 18 pages, for Northgate Tax & Advisory, PC - a made-up 6-person firm. Every person, device and vendor in it is invented, apart from CPA Hub itself, which sits in its vendor list the way it sits in every real firm's. It was not drawn for this page: it is what CPA Hub's own export prints for that firm, with SAMPLE across every sheet.
What is in it
Page numbers are the ones printed at the foot of each sheet. After the signature page come the attachments - the designated official, the retention schedule, the device inventory and the risk register - drawn from the firm's records rather than typed into the plan.
- Program Summary - the ten facts an examiner asks for, with the open findings firstpage 1
- About this documentpage 2
- Designated Security Coordinatorpage 3
- Risk Assessmentpage 4
- Multi-Factor Authentication & Access Policypage 5
- Encryption Policypage 6
- Incident Response Planpage 7
- Employee Security Trainingpage 8
- Third-Party Vendor Managementpage 9
- Data Retention & Disposalpage 10
- Responsible AI Use - Circular 230, as read in IRS OPR Issue Number 2026-19page 11
- Acknowledgment & approval (signature page)page 13
Page one: what an examiner asks for first
The Program Summary on page 1 lists the ten facts an FTC examiner or an insurer asks about, each against the paragraph of 16 CFR 314 it answers. For this firm, 8 of 10 are answered and current and 2 are open, and the open ones lead the page: 1 device holds client data and is not encrypted — starting with Front-desk scanner PC. A plan that says everything is fine is not a plan anyone believes; this one names its own gaps.
The new part: Responsible AI Use (Circular 230)
On June 24, 2026 the IRS Office of Professional Responsibility issued Issue Number 2026-19, "Introductory Guidelines for Responsible AI Use in Federal Tax Practice". Introductory guidelines. They create no new rule; each point rests on an existing Circular 230 duty that OPR enforces. The sample's section on page 11 is written to it. It sits after the eight FTC sections and says plainly that it is not one of them, so nothing in it counts toward an FTC finding. What Issue 2026-19 asks of a firm, in plain English.
Where the made-up firm stands - 5 of 7 answered and current:
| Status | What is checked | Authority |
|---|---|---|
| OK | Written AI policy Written and saved in this year's WISP | 31 CFR 10.36 |
| OK | Human review of AI output Every AI output is reviewed by a person before it reaches a client or the IRS | 31 CFR 10.22, 10.37 |
| OK | AI tools listed 3 AI tools in the service-provider register | 31 CFR 10.36 |
| OK | Client data in AI tools 3 approved for client data, 0 not | IRC 6713, 7216(a) |
| NOT RECORDED | Vendor training on firm data NOT RECORDED for 1 approved tool | 31 CFR 10.36; IRC 7216(a) |
| OPEN | Staff trained on AI 5 of 6 passed the AI module | 31 CFR 10.35, 10.36 |
| OK | Record of AI use The firm keeps a record of where AI was used and how it was checked | 31 CFR 10.36 |
The firm's AI policy, as it appears in the sample
Northgate follows IRS Office of Professional Responsibility Issue Number 2026-19, "Introductory Guidelines for Responsible AI Use in Federal Tax Practice" (June 24, 2026). Those guidelines create no new rule; each point below rests on an existing Circular 230 duty.
Approved tools. Only AI tools the Data Security Coordinator has approved are used for firm work. Each is listed, with its decisions, in the firm's service-provider register (31 CFR 10.36). Personal and free AI accounts are not used for firm work.
Client data. Client information goes only into AI tools approved for it, under business terms that protect confidentiality, and never into a public or personal account (IRC 6713, 7216(a)).
Human review. Everything an AI tool produces is a draft. Before it reaches a client or the IRS, a preparer checks the facts, the calculations and every citation (31 CFR 10.22). Written advice never rests on AI output that has not been independently verified (31 CFR 10.37).
Fees. Bills reflect the time the work actually took; the firm does not charge for manual time an AI tool replaced (31 CFR 10.27(a)).
Training. Everyone who uses AI completes the firm's AI module before using it and every year after (31 CFR 10.35, 10.36).
Records. Where AI was used on a return and how it was checked is noted in the workpapers. CPA Hub records its own AI use automatically.
Errors and disclosures. A wrong AI output that reached a client, or client information entered into an unapproved tool, is reported to the Data Security Coordinator the same day, corrected, and handled under the incident response plan where needed.
This policy is reviewed every year with the rest of the plan.
Frequently asked questions
Sources
- IRS Office of Professional Responsibility, Issue Number 2026-19: Introductory Guidelines for Responsible AI Use in Federal Tax Practice (June 24, 2026)
- IRS Publication 5708, Creating a Written Information Security Plan: irs.gov/pub/irs-pdf/p5708.pdf
- FTC Safeguards Rule, 16 CFR Part 314: ecfr.gov
General information, not legal advice. The sample describes an invented firm and is not a plan for yours.