Yes, lawyers can use ChatGPT and similar tools ethically, but the existing rules still apply in full: you need to understand what the tool can and cannot do, protect client confidences, supervise how your staff use it, verify everything it produces, and bill fairly. In practice that means a firm-approved account, no client confidences without proper safeguards or consent, and a lawyer reading every output, the habit you practice in Answer Case-Update Emails Without Predicting the Outcome.
This page summarizes published guidance in plain words. It is not legal or ethics advice. Your state's rules and bar opinions control, so check them before you adopt any tool.
What does the ABA say about generative AI?
In July 2024 the ABA issued Formal Opinion 512, its first ethics guidance on lawyers' use of generative AI tools. It does not create new rules. It applies the Model Rules you already know:
- Competence (Rule 1.1). You do not need to be an AI expert, but you need a reasonable understanding of the tool's capabilities and limits, and you must check its output to the degree the task requires.
- Confidentiality (Rule 1.6). Before putting client information into a tool that may learn from it, the opinion says informed client consent is generally needed, and boilerplate in an engagement letter is not enough. Using a tool for general ideas with no client information is different.
- Communication (Rule 1.4). Whether to tell a client you use AI depends on the facts, such as when the client asks or when it affects fees or significant decisions.
- Supervision (Rules 5.1 and 5.3). Partners and supervising lawyers need policies and training so associates, paralegals and staff use AI properly.
- Fees (Rule 1.5). Bill for the time actually spent, including reviewing output. In most cases you cannot bill clients for learning the tool.
Many state bars have issued their own AI guidance as well. Read yours, because states differ on details like disclosure and consent.
Why do citations get so much attention?
Because general chatbots can invent cases, quotes and holdings that look real. Courts have sanctioned lawyers who filed briefs with citations an AI made up. A general assistant is not a research database. Every case, statute and quotation must be checked in a trusted source before it goes into anything you file or send. Legal-specific tools like CoCounsel, Thomson Reuters' AI legal assistant for research, review and drafting, are built for legal work, but the duty to verify stays with you.
What safeguards should a firm put in place?
- Approve specific tools and plans. Business plans that do not train on your data by default, with admin control over accounts.
- Write a one-page AI policy. What may go in, what never goes in, who reviews, how to report a mistake.
- Check the settings. Training off where it is a setting, memory managed, shared links reviewed.
- Decide on client consent and disclosure for each type of use, with your state's guidance in hand.
- Require lawyer review for anything that reaches a client, a court or opposing counsel.
- Train staff, especially paralegals and assistants who draft client communications.
ChatGPT: Keep Client Data Safe covers the settings and a one-page team policy, and Use AI at Work Without Leaking Client Data covers what to take out before you paste.
What does this look like in a real lesson?
In Answer Case-Update Emails Without Predicting the Outcome Carla, a flower shop owner, is suing an event venue that never paid her 8,400 dollars. She wants a court date, a promise she will win and permission to text screenshots. At Harbor Street Law you paste the thread into ChatGPT with the firm's facts: what has been filed, the new deadline, what a missed deadline does and does not mean, the one court date, what the firm cannot predict and the no-texting policy. The first draft sounds reassuring, but when you ask for a table of every claim and the fact behind it, three are unsupported: an automatic win, a "court date" that is not one, and a prediction about getting paid. You use the safer wording, add a line about keeping the original texts, and send. If your firm has an attorney approve client updates, they read it too.
That is competence and communication in practice: the tool drafts, and the lawyer makes sure nothing predicts an outcome or gives advice the facts do not support. The legal information vs legal advice guide and ChatGPT for law firms go deeper.
What mistakes should lawyers avoid?
- Filing anything with an unchecked citation.
- Putting client confidences into a personal account.
- Relying on engagement-letter boilerplate for consent.
- Billing the time AI saved as if you had spent it.
- No policy for staff, so each person makes up their own rules.
Key takeaways
- The Model Rules already cover AI: competence, confidentiality, communication, supervision and fees.
- Protect client information with approved business tools and, where needed, informed consent.
- Verify every citation and fact in a trusted source.
- A lawyer reviews anything that reaches a client or a court.
Frequently asked questions
- Do I have to tell clients I use ChatGPT?
It depends on the use and your jurisdiction. ABA Formal Opinion 512 treats disclosure as fact-specific, for example when a client asks or when AI use affects fees. Check your state bar's guidance.
- Is ChatGPT Enterprise safe enough for client files?
A business plan that does not train on your data is a strong start, but it does not settle consent, supervision or verification. Those are still your duties.
- Can paralegals use ChatGPT at my firm?
Yes, under your supervision and your firm's policy. See where a paralegal should start with AI.