Is It Safe for a Lawyer to Put Client Files in ChatGPT?
Short answer: On a personal ChatGPT account, usually not: OpenAI says that on Free, Plus and Pro "data sharing is enabled by default", which means your chats can help train its models unless you switch that off. A business plan is a different question. ChatGPT Business and Enterprise do not train on your data by default, but they still keep your chats, so the ethics rules still ask you to read the terms, know who can see the file and decide whether the client must consent first.
What the ethics rules ask before a file goes in
Nothing in the ABA Model Rules bans a particular AI tool. ABA Formal Opinion 512 asks for something more practical: that you understand the tool well enough to protect the client's information under Rule 1.6. For tools that can learn from what you give them, it says "a client's informed consent is required prior to inputting information relating to the representation into such a GAI tool" (p. 7). It also tells lawyers to "learn who has access to the information that the lawyer inputs into the tool" (p. 7), and to find out whether the tool "retains information submitted by the lawyer before and after the discontinuation of services" (p. 11) (ABA Formal Opinion 512).
Your state's rules control, and several states have their own guidance. See the state-by-state guide.
Which ChatGPT are you using?
"ChatGPT" covers several plans with different terms. Each quote below is from OpenAI's own pages, read on November 13, 2026.
| Plan | Trains on your chats? | Keeps your chats? | Who at OpenAI can see stored chats |
|---|---|---|---|
| Free, Plus, Pro (personal) | Yes by default: "data sharing is enabled by default", and you can turn off "Improve the model for everyone" (data controls) | Yes, in your history until you delete them. A temporary chat "may be retained for up to 30 days for safety purposes" | Not set out on the page we read |
| ChatGPT Business | No by default: "We don't train our models on your organization's data by default." (OpenAI) | "Your workspace admins can control how long your data is retained. Any deleted or unsaved conversations are removed from our systems within 30 days, unless longer retention is required by law, or is reasonably necessary to protect our services or any third party from harm." | "authorized employees that require access for engineering support, investigating potential platform abuse, and legal compliance" and "specialized third-party contractors" (enterprise privacy) |
| ChatGPT Enterprise | No by default, as above | "Your workspace admins control how long your data is retained. Any deleted conversations are removed from our systems within 30 days, unless we are legally required to retain them." | "Authorized OpenAI employees will only ever access your conversations for the purposes of resolving incidents, recovering end user conversations with your explicit permission, or where required by applicable law." |
Two things follow for a lawyer, in our reading. A personal account with training left on looks like the self-learning tool the opinion describes, so the consent question comes first. And on every plan, stored chats are records OpenAI holds, which legal process can reach. The 2025 preservation order in NYT v. OpenAI did not cover ChatGPT Enterprise, but it shows how a court can reach stored chats. Our page on confidentiality and privilege covers that order and the Heppner ruling.
How other tools lawyers use compare
The same three questions apply to any tool: does it train on your input, does it keep it, and who can see it.
| Tool | Training, in its own words | What it keeps |
|---|---|---|
| Claude Team and Enterprise (Anthropic) | "No model training on your content by default" (pricing) | "we retain your chats and coding sessions in the product"; deleted chats leave back-end storage "within 30 days" |
| Microsoft 365 Copilot | "Prompts, responses, and data accessed through Microsoft Graph aren't used to train foundation LLMs" | "The stored data includes the user's prompt and Copilot's response"; admins set retention in Microsoft Purview |
| Harvey | "We don't use inputs, outputs, or uploaded documents to train underlying models." | What you upload, under "retention policies" your firm sets |
| Spellbook | Not set out on the security page we read | Zero data retention at "both OpenAI and Anthropic"; "Other third party services are used that may store or process customer information" |
| Faradex | 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 | Files and conversations are held in memory and are gone when the Workspace is reset |
A longer comparison, with pricing and setup, is on AI providers for law firms.
Five steps before you upload a client file
- Use a firm account, not a personal one. Personal plans are where training is on by default.
- Read the plan's data terms yourself, or have someone qualified read them, as the opinion expects. Note who can see stored chats and for how long.
- Decide whether you need the client's informed consent. If the tool can learn from the file, the opinion says yes. Boilerplate in an engagement letter does not count.
- Strip what the task does not need. Names, account numbers and medical details are often unnecessary for a summary or a first draft.
- Write down what you used and why. A short note in the file shows the judgment you made if anyone asks later.
The full pre-use list, with the rule behind each item, is on the checklist page.
This page is general information, not legal advice. Check your state's rules and each vendor's current terms.