---
title: "Using ChatGPT for legal work: safe uses, settings and limits"
url: https://notionforlawyers.com/ai/chatgpt
updated: 2026-10-17
author: Raghav Handa
---

# Using ChatGPT for legal work: safe uses, settings and limits

ChatGPT can help lawyers with drafting structure, summaries of non-confidential text, checklists and plain-language rewrites. It should not be treated as a legal research source, and client information should only go in if your firm has approved the specific plan after reading the vendor's current data terms. Always verify output.

## Key takeaways

- Safe starter tasks use public or anonymised text: outlines, checklists, rewrites and brainstorming.
- Data controls differ by plan (consumer, team, business, enterprise). Read the current terms for the plan you actually use.
- Never rely on unverified citations; courts have sanctioned lawyers who filed fabricated cases.
- Write an internal AI-use policy covering what may be entered and who reviews output.

## What ChatGPT is good for in a law practice

General information, not legal advice. Check your own bar or regulator's guidance and your client engagement terms before putting any client information into an AI tool.

ChatGPT is a general-purpose assistant from OpenAI. It is strongest at language tasks: restructuring a rambling note, drafting a first version of a routine letter, producing a checklist, or explaining a concept in plain English. It is weaker where precision and currency matter: statutes, case law, deadlines and jurisdiction-specific rules.

- Outlines and first drafts of non-confidential documents
- Summaries of public material
- Brainstorming arguments or questions to ask a client
- Reformatting data into tables
- Marketing copy and internal training material


## Settings and data controls to check

Vendors change these settings and terms regularly, so this is a list of questions, not a statement of current policy. Check OpenAI's current terms and privacy documentation for the exact plan you use.

| Question to answer | Where to look | Why it matters |
| --- | --- | --- |
| Is my content used to train models by default? | Plan terms and data-controls settings | Consumer and business plans have historically differed |
| Can I turn off chat history or training use? | Account data-controls settings | Reduces retention but does not remove the confidentiality duty |
| How long is data retained? | Privacy policy and plan terms | Affects your file-retention and deletion obligations |
| Is there a data processing agreement? | Vendor's business or enterprise terms | Often needed for regulated data and client requirements |
| Who can see shared chats? | Workspace admin settings | Shared links can expose content |


## A cautious workflow

1. Decide the data class: Public, anonymised, or confidential. Only the first two go into an unapproved tool.
2. Anonymise properly: Remove names, addresses, case numbers and any unique fact that identifies the client.
3. Prompt for structure, not authority: Ask for outlines, questions and drafts; ask it to flag assumptions.
4. Verify: Check every legal statement and citation in a primary source.
5. Record: Keep a note of what tool was used for what, as your policy requires.

For ready-made prompts, see the prompt library. For a Notion-native option, see Notion AI.

- AI prompt library: https://notionforlawyers.com/ai/prompts
- Notion AI for lawyers: https://notionforlawyers.com/ai/notion-ai


## Mistakes to avoid

Treating the output as research. Lawyers in several jurisdictions have been sanctioned for filing briefs containing fabricated citations produced by AI tools.

Pasting client documents into a personal account. Personal consumer accounts are the hardest to govern and audit.

Assuming a setting equals compliance. Turning off training use does not by itself satisfy confidentiality, retention or client-consent obligations.

Skipping supervision. Under professional-conduct rules a lawyer remains responsible for work product, including work assisted by tools. ABA Formal Opinion 512 (July 2024) addresses lawyers' use of generative AI; check your own regulator's guidance too.


## FAQ

### Is ChatGPT safe for confidential client data?

It depends on the plan, the vendor's current terms and your firm's policy. Without clear approval, assume confidential data should not be entered.

### Can ChatGPT replace legal research tools?

No. It can help frame a research question, but authorities should come from a reliable legal database and be read directly.

### Does turning off chat history make it safe?

It may reduce retention or training use depending on the plan, but you still need to meet your confidentiality duties. Check the current terms.

### How do I compare ChatGPT with Claude and Gemini?

Use the same anonymised test task in each and compare output, then compare data terms. See our Claude and Gemini pages.
