AI chatbots like ChatGPT, and similar tools, are increasingly used by lawyers for help with drafting, summarising and other tasks. Used well, they can save time; used carelessly, they can cause serious errors, breaches and professional harm. ChatGPT itself acknowledges that it can produce incorrect or misleading output, including fabricated citations and references, which is why important information should be independently verified. The difference is following sound best practices. Let me give you the essential best practices for using ChatGPT or similar AI in legal work responsibly.
Quick answer
The essential best practices for using ChatGPT or similar AI in legal work are: verify everything it produces, because AI can be confidently wrong and fabricate cases, facts and citations, so never rely on unverified output; protect confidentiality, by not inputting confidential or client information into such tools without proper safeguards and compliance with your obligations; use it for suitable tasks only, as an assistant for things like first drafts, summaries or support, not as a source of legal truth or a substitute for judgment; keep full professional responsibility, because you are responsible for the accuracy, quality and propriety of your work, not the tool; use good prompts and review critically, because clear prompts help, but output still requires careful review; and comply with all applicable rules, including your professional and ethical obligations and any firm or professional-body rules on AI use. OpenAI itself recommends treating ChatGPT as a first draft rather than a final source and double-checking important information.
Follow these practices and AI can be a helpful assistant. Ignore them and it can cause serious problems. The keys are: verify, protect confidentiality, use appropriately, keep responsibility, review critically and comply. Because the technology and rules evolve quickly, stay current and verify the position. This is general guidance, not legal or professional advice.
The essential best practices
1. Verify everything
AI can produce confident but false or fabricated output, including made-up cases and citations. ChatGPT’s own guidance specifically warns that it may generate incorrect facts, fabricated citations or references, and advises users to verify important information from reliable sources.
Never use its output without verifying accuracy and truth yourself. This is the single most important practice. See “AI Hallucinations in Legal Work.”
There have already been real examples of lawyers relying on fabricated AI-generated authorities. In Mata v. Avianca Inc., a US federal court sanctioned lawyers after they submitted non-existent judicial opinions and fake citations generated by ChatGPT.
2. Protect confidentiality
Do not input confidential or client information into AI tools without proper safeguards and compliance. Assume caution about what you put into such tools, particularly where the material contains privileged, confidential, personal or commercially sensitive information. See “Confidentiality and Client Data Risks When Using AI.”
3. Use it for suitable tasks only
Use AI as an assistant for suitable tasks such as first drafts, summaries, brainstorming and other preparatory support, not as a source of legal truth or a substitute for your judgment and expertise.
It can help with the work, but you still have to think through the work.
4. Keep full professional responsibility
You remain fully responsible for your work. The tool is an aid; the responsibility, judgment and final work are yours.
That responsibility does not disappear because a draft, summary or research lead came from an AI system. The lawyer remains the person who must decide whether the final work is accurate, appropriate and fit for its purpose.
5. Use good prompts and review critically
Clear, well-crafted prompts can help you get more useful results, but prompting is not a substitute for review. Always examine the output critically, check its reasoning and verify the important parts rather than accepting it uncritically. OpenAI similarly recommends keeping a human in the loop for important work and double-checking critical facts against trusted sources.
See “How to Write Effective Prompts as a Lawyer.”
6. Comply with all applicable rules
Ensure your use of AI is consistent with your professional and ethical obligations and any applicable firm or professional-body rules on AI. If your workplace has an AI policy, follow it.
AI use should fit within the obligations that already govern your legal work. The technology does not remove those obligations. See “What Law Firm AI Policies Typically Cover” and “AI Ethics for Lawyers.”
Why these matter so much for lawyers
For lawyers, the stakes of getting this wrong are high. Relying on fabricated AI output can lead to serious errors. In Mata v. Avianca, for example, lawyers submitted fake case authorities generated by ChatGPT and were sanctioned by the court.
Confidentiality failures can harm clients and your professional standing, while failing to retain responsibility for AI-assisted work can create serious professional problems. So these best practices are not optional niceties. They are essential to using AI safely and professionally.
Follow them consistently, and AI can become a useful tool rather than a liability.
YLCC ACTION STEP: Before using ChatGPT or similar AI in any legal work, commit to the essentials: verify everything it produces, never input confidential or client information without appropriate safeguards, use it only for suitable tasks, keep full professional responsibility, review output critically, and comply with your obligations and any applicable rules. Make these habits automatic. Responsible use is what makes AI a helpful tool rather than a risk.
If this is you
If you use AI casually without much thought: Adopt the best practices deliberately, especially verifying output and protecting confidentiality. Casual, careless use is exactly where serious problems can arise.
If you have relied on AI output before: Going forward, verify everything. AI can fabricate cases and facts confidently, so never submit or rely on unverified AI output. There have been real professional consequences for doing so.
If your firm or context has AI rules: Follow them. Ensure your use complies with your firm’s or profession’s rules on AI, as well as your general professional obligations. See “What Law Firm AI Policies Typically Cover.”
PLEASE DON’T DO THIS: Please don’t rely on or submit AI output without verifying it. It can fabricate cases, facts and citations, with real professional consequences. Please don’t input confidential or client information into AI tools without proper safeguards. And please don’t treat the tool as responsible for your work or ignore applicable rules. Verify, protect confidentiality, keep responsibility, review critically and comply.
FAQs
- What are the best practices for using ChatGPT in legal work? Verify everything it produces, protect confidentiality by not putting client data into tools without appropriate safeguards, use it for suitable tasks only, keep full professional responsibility, use good prompts and review critically, and comply with all applicable rules.
- Can I trust ChatGPT’s legal output? No, not without verifying it. AI can be confidently wrong and can fabricate cases, facts and citations. Always verify accuracy and truth yourself.
- Can I put client information into ChatGPT? Not without proper safeguards and compliance. Doing so can create serious confidentiality and data risks. Protect confidentiality and follow your professional obligations.
- Who is responsible if the AI is wrong? You are. You remain fully responsible for your work; the tool is not. Verify everything before relying on or submitting AI-assisted work.
- What if my firm has AI rules? Follow them, along with your general professional and ethical obligations. Ensure your AI use is consistent with all applicable rules.
