Short answer: Neither the SEC nor FINRA has adopted a rule written specifically for AI, so "SEC compliant AI" describes your firm's process, not a product. Three things decide whether an AI tool fits that process: you can oversee it as a Regulation S-P service provider, you know where anything that counts as a record ends up, and your written policies say who may use it on client data. Microsoft 365 Copilot, ChatGPT Enterprise and Claude Enterprise each handle those points differently, mainly in what they store and who controls retention.
No rule names an approved AI tool
FINRA's notice on generative AI says it "does not create new legal or regulatory requirements" (FINRA Regulatory Notice 24-09). The SEC's examination priorities for fiscal 2027 say staff will look at "oversight of service providers" under Regulation S-P, and the document adds that it "has no legal force or effect" (SEC). So when a vendor says its tool is SEC compliant, read that as a claim that the tool can fit into a compliant program. The program is yours. What the regulators have said so far is on the SEC and FINRA page.
Test 1: can you oversee it as a service provider?
Regulation S-P defines a service provider as "any person or entity that receives, maintains, processes, or otherwise is permitted access to customer information through its provision of services directly to a covered institution" (17 CFR 248.30). An AI tool that sees a client statement or a financial plan fits those words. Your policies then have to be designed to require the vendor to tell you of a breach within 72 hours of becoming aware of it. The details are on Regulation S-P and AI.
The practical question for any AI vendor: what does it hold, and would it tell you within 72 hours if that were exposed? A tool that stores less gives that question less to cover.
Test 2: do you know where the records go?
The books-and-records rule cares about what a record says, not which tool wrote it. A final client email about advice is a record whether or not AI drafted it. Two patterns fit the rule:
- Keep AI history inside a system you already archive. Microsoft says Copilot's "stored data includes the user's prompt and Copilot's response", and that admins "can also use Microsoft Purview to set retention policies" for it (Microsoft Learn). That suits a firm whose archive already runs on Microsoft 365.
- Keep the final work product, not the AI chat. The advisor saves the finished email or memo to the CRM or archive, and the AI tool keeps as little as possible. That suits a firm that does not want a second store of client data to oversee.
Record-keeping and AI covers which rules apply.
Test 3: will your written policies say who may use it?
Neither pattern works without a short written policy: which tool, which plan, which kinds of client information, and who reviews output before it reaches a client. Start by ruling out personal AI accounts for client work. OpenAI says that on ChatGPT Free, Plus and Pro, "data sharing is enabled by default" for model training (OpenAI).
How four options answer the three tests
We read every vendor page linked below on November 13, 2026, and quote it directly. We do not score or rank these tools.
| What the vendor stores | Who sets retention | Training | |
|---|---|---|---|
| Microsoft 365 Copilot | Prompts and responses, as Microsoft 365 content | Your admins, in Microsoft Purview | "Prompts, responses, and data accessed through Microsoft Graph aren't used to train foundation LLMs" |
| ChatGPT Enterprise | Conversations; deleted ones "removed from our systems within 30 days, unless we are legally required to retain them" | "Your workspace admins control how long your data is retained." | "We don't train our models on your organization's data by default." (OpenAI) |
| Claude Enterprise | Chats and projects; "by default, projects are retained indefinitely" | Your admins; "The minimum retention period is 30 days" | "No model training on your content by default" (Claude pricing) |
| Faradex | Files and conversations are held in memory and are gone when the Workspace is reset | Faradex logs no content. Download any output you want to keep, and file it in your firm's own archive. | Currently Opus via AWS Bedrock, more information is available on our website. AWS does not store prompts or responses, and model providers have no access to them |
Claude Enterprise is sold at "US$20/seat/month, billed annually, 20 seat minimum", plus usage. ChatGPT Enterprise has no public price. Copilot needs a Microsoft 365 plan. Advisor-specific tools such as Jump and Zeplyn, which sync notes to your CRM and archive, are compared on Private AI for wealth management firms.
Questions to put to any AI vendor before client data goes in
- Which of our data do you store, where, and for how long? Can we shorten it?
- Who on your side can read stored content, and for what reasons?
- Will you notify us within 72 hours of a breach that touches our clients' information, in the contract?
- Do you, or your model provider, train on our data? Is that in the contract or only on a web page?
- If our policies require us to keep an AI-assisted record, how do we export it to our archive?
This guide is general information, not legal, compliance or investment advice. Your compliance lead decides how these rules apply to your firm.