Vendor note. Verlo makes AI software for financial advisors. This template is vendor-neutral and does not mention our product. It is a starting point, not legal advice, and the sample language is ours rather than rule text. Have your compliance officer or counsel adapt it to your firm, your custodians and your state requirements before you adopt it. If something here is out of date, tell us at support@verlo.finance. If you are also choosing tools, our guide to the best AI software for financial advisors compares them.
The short answer. A workable AI policy is two to four pages. It names the approved tools, sorts data into tiers, requires a person to review AI output, applies your marketing review to AI-assisted content, says what you keep, ties every tool to the vendor due diligence Regulation S-P requires, sets the incident steps, and schedules an annual review. This template covers all ten.
SEC-registered advisers must have written compliance policies reasonably designed to prevent violations of the Advisers Act, review them at least annually and have a chief compliance officer to run them. An AI-use policy is how most firms show they have applied that to AI. We found no RIA-specific template with sample language, so this one gives you wording to start from. Use it as the AI section of your compliance policy manual, or as a standalone policy your manual points to.
Download the policy template (Word) and the approved-tool register (spreadsheet).
How to use this template
Tailor the bracketed items. Wherever you see braces, such as {Firm Name}, choose your own wording.
Make the decisions listed under each section. Each one has a few choices only you can make.
Have counsel read it, especially the records, incident and marketing sections.
Adopt it, train the team and keep it with your compliance manual, alongside the approved-tool register and your vendor files.
The template, section by section
1. Purpose and scope
The policy needs to say what it covers. Include AI features inside tools you already use, such as a CRM, an email client or a meeting assistant, not only standalone chatbots.
Sample language. {Firm Name} (the “Firm”) permits the use of artificial intelligence tools to support its work, subject to this policy. This policy applies to all supervised persons and to any AI tool, assistant, plug-in or feature used for Firm business, whether provided by the Firm or acquired personally. It supports the Firm’s obligations under the Investment Advisers Act of 1940 and the rules under it, including Regulation S-P, Rule 204-2, Rule 206(4)-1 and Rule 206(4)-7.
Decide: whether AI features built into existing tools are in scope (we recommend yes), and whether the policy also covers contractors.
2. Roles and responsibilities
Rule 206(4)-7 requires a designated chief compliance officer. Name who owns this policy and who supervises day-to-day use.
Sample language. The Chief Compliance Officer ({name or title}) administers this policy: approving AI tools, maintaining the approved tool register and vendor files, and receiving reports of exceptions and incidents. Supervisors are responsible for use in their areas. Every supervised person is responsible for using only approved tools, following the data rules in Section 4, and reporting a suspected incident immediately.
Decide: who is the backup when the CCO is unavailable, and who approves exceptions.
3. Approved tools and accounts
Make the approved list the control. Anything not on it is not approved for client information.
Sample language. Supervised persons may use only AI tools on the Firm’s approved tool register, and only through Firm-managed accounts on the plan the register specifies. Personal or free accounts may not be used for any Firm business involving client information. A tool is added to the register only after the vendor due-diligence process in Section 8 is complete and the CCO has approved it in writing. For each tool the register records the vendor, plan, approved uses, data tiers permitted, systems connected, owner, date approved and next review date.
Decide: which plans qualify (for example, business or enterprise plans with no training on your data by default), and how staff request a new tool.
4. Data rules
Sort data into tiers so staff do not have to guess. Redacting a name is not a safe harbor: the remaining details can still identify a household.
Sample language. AI use is governed by three data tiers. Tier A: no client information; any approved tool may be used. Tier B: client information; only tools the CCO has approved for client information may be used. Tier C: never entered into any AI tool, including passwords and credentials, full Social Security and account numbers, and {other categories}. Removing a client’s name does not make information Tier A. If in doubt, treat information as client information and ask the CCO.
Decide: your Tier C list, and whether any Tier B tool may receive tax returns, estate documents or recordings.
5. Human review and supervision
The tool is not accountable; the person is. Say so, and say what AI may not decide.
Sample language. AI output is a draft. A supervised person must review all AI-generated content before it is sent to a client, filed, published or relied on, and is responsible for its accuracy. Names, numbers, dates and rules must be checked against the source. AI may not be used to make or finalize investment recommendations, best-interest determinations or tax or legal conclusions without documented review by a qualified person. Supervisors may sample AI use as part of the Firm’s supervisory procedures.
Decide: what documented review looks like for each activity, and how you sample use.
6. Marketing and client communications
The Marketing Rule applies to an advertisement however it was drafted, and it also applies to what you say about your own use of AI.
Sample language. AI-assisted advertisements, newsletters, social media posts and website content are subject to the Firm’s advertising review under Rule 206(4)-1 before use, regardless of how they were drafted. AI may not be used to create or alter performance information, testimonials or endorsements without the review required by the Firm’s marketing procedures. Statements about the Firm’s own use of AI, in marketing or regulatory filings, must be accurate, specific and approved by the CCO.
Decide: who reviews AI-assisted content, and whether any categories are prohibited outright.
7. Records and retention
Rule 204-2 sets the retention period for records it requires. Decide which AI outputs and transcripts fall under it, by category of use, and where they are kept. We found no SEC guidance on AI transcripts and summaries, so make an explicit decision.
Sample language. The Firm retains AI-related records as follows: {categories, for example client communications drafted with AI, as sent; meeting notes and summaries filed to the client record; transcripts or recordings, if kept, for {period}; prompts and outputs relied on for advice or compliance decisions}. Records required by Rule 204-2 are kept for at least five years from the end of the fiscal year of the last entry, the first two years in an appropriate Firm office. Supervised persons must file AI-generated records the Firm requires in {system}. A vendor’s own copy is not the Firm’s record unless the register says so.
Decide: which meeting categories you keep transcripts for, how you export and search records, and what happens when a vendor account is closed.
8. Vendor due diligence and monitoring
Regulation S-P requires written oversight of service providers, through due diligence and monitoring. This section is where the policy meets your vendor file.
Sample language. Before a tool is used with client information, the CCO completes and documents due diligence on the vendor, covering at least: the data handled and where it is stored; training and retention terms; model providers and subprocessors; security attestations; notification of a breach, including notice to the Firm no later than 72 hours after the vendor becomes aware of a breach of a system holding customer information; access controls and logging; and exit and deletion terms. Material terms are put in writing. Vendors that handle client information are re-reviewed at least annually and on any material change. The vendor file is kept as a Firm record.
Decide: your evidence standard (for example, a current SOC 2 Type 2 report), and how you track re-review dates. Our due-diligence checklist gives the questions, and this guide lists what to keep in the file.
9. Incident response and escalation
Tie this to the incident response program Regulation S-P requires. The policy should say who is told, how fast and what is recorded.
Sample language. A supervised person who suspects unauthorized access to or use of customer information, or who learns of a vendor breach, must notify the CCO immediately. On receipt of a vendor notice, the CCO initiates the Firm’s incident response program under its Regulation S-P procedures, assesses the nature and scope of the incident, takes steps to contain it, determines whether notice to affected individuals is required, and documents each step. Where notice is required, it is given as soon as practicable and no later than 30 days after the Firm becomes aware of the incident, unless an exception applies.
Decide: who contacts clients, whether any vendor may notify on the Firm’s behalf under a written agreement, and where incident records live. See our Regulation S-P guide.
10. Training, exceptions and annual review
Rule 206(4)-7 requires a review of your policies’ adequacy and effectiveness at least annually. Put it on the calendar.
Sample language. All supervised persons receive training on this policy at onboarding and at least annually. Exceptions require the CCO’s prior written approval and are logged. The CCO reviews this policy, the approved tool register and the vendor files at least annually, and whenever the Firm adopts a new category of AI tool, a vendor materially changes its terms or models, or the regulatory position changes, and documents each review.
Decide: the training format, how exceptions expire, and the month your annual review happens.
Approved tool register
Keep one register for every AI tool the Firm approves, and review it with the policy. Here is a starting layout, with an example row:
Tool | Vendor | Plan or product | Approved uses | Data tiers permitted | Systems connected | Owner | Date approved | Due-diligence file | Training and retention terms | 72-hour breach notice in contract | Next review |
|---|---|---|---|---|---|---|---|---|---|---|---|
Example notetaker | Example Vendor Inc. | Business plan | Client meeting notes | A and B | CRM, calendar | Operations lead | Date | Link to file | No training; retention set to policy | Yes | Date |
Download this table as a spreadsheet (CSV)
Incident quick checklist
Tell the CCO immediately, and record the time you became aware.
Contain the incident: revoke access, disconnect the tool, preserve logs.
Assess what customer information was involved, and whether any of it is sensitive customer information.
Decide, with counsel, whether notice to affected individuals is required, and document the basis.
If notice is required, send it as soon as practicable and no later than 30 days after you became aware.
Keep the documentation, the notice and any vendor agreement in the incident file.
Review what failed, update the policy and the vendor file, and brief the team.
Frequently asked questions
Do I need an AI policy?
No rule uses that phrase, but Rule 206(4)-7 requires written policies and procedures reasonably designed to prevent violations, reviewed at least annually, and Regulation S-P requires written oversight of service providers. An AI-use policy is how most firms apply both to AI.
How long should an RIA’s AI policy be?
Two to four pages is workable. A short policy that staff read and follow beats a long one that does not. Put detail in the register, the vendor files and the playbook for each tool.
Should the policy ban personal AI accounts?
We recommend barring personal and free accounts for any client information, because they generally do not give you written commitments on training, retention and breach notice. See Can financial advisors use ChatGPT or Claude with client data?
How do we add AI to our compliance policies and procedures manual?
Add it as its own section of the manual, or as a standalone policy the manual references, rather than a memo staff never see. Then connect it to the sections it touches: books and records (what AI output you keep), vendor oversight (the due diligence Regulation S-P requires), marketing review (AI-assisted content), supervision (who reviews AI output) and incident response. The ten sections of this template are written so each one can sit next to the matching part of your manual. Record the adoption date and include the AI section in your annual review. Have counsel read it before you adopt it.
Who should approve the policy?
The CCO administers it, and senior management should approve it, as you do for the rest of your compliance manual. Counsel should review it before adoption.
How often should we update it?
At least annually, and when you add a new category of tool, a vendor changes its terms or models, or the rules change. The sample language builds that in.
Go deeper
Updates to this guide
November 17, 2026: First published.
November 19, 2026: Added a question on adding the policy to your compliance policies and procedures manual.
Sources
Checked November 6, 2026: 17 CFR 275.206(4)-7-7), 17 CFR 275.206(4)-1-1), 17 CFR 275.204-2, 17 CFR 248.30.