Can lawyers use ChatGPT? The confidentiality rules, straight.
Short answer: yes, lawyers can use AI (the ABA said so explicitly), but not with client information in consumer tools. The line isn't the technology; it's Rule 1.6. Here's what the ethics guidance actually says, and a safe/unsafe table you can hand to everyone in the office.
What the ethics guidance actually says
ABA Formal Opinion 512 (July 2024) permits lawyers to use generative AI, under the ordinary duties that already govern everything else: competence, confidentiality, supervision, candor, and reasonable fees. It did not create an AI exception in either direction. The Illinois Supreme Court Commission's analysis walks through the same rules: the Rule 1.6 duty to make "reasonable efforts to prevent the inadvertent or unauthorized disclosure of… information relating to the representation of a client", and notes that consumer ChatGPT's own terms say non-API content may be used to train models. Put those two together and the conclusion writes itself: typing client facts into a consumer chatbot risks breaching the duty of confidentiality and endangers privilege. Not because AI is special, but because you handed client information to a vendor with no confidentiality obligation to you.
The safe / unsafe table
| Use | Verdict | Why |
|---|---|---|
| Drafting templates, letters, and marketing with no client identifiers | Safe | No confidential information leaves the office. |
| Summarizing public documents, statutes, opinions | Safe | Public in, summary out. Verify before relying: competence still applies. |
| Intake automation (forms, scheduling, conflict-check workflows) | Safe with the right tools | Business-grade tools with no-training, zero-retention terms, never consumer tiers. |
| Client matters in business-grade AI under written no-training terms | Permitted with care | This is the Opinion 512 path: vendor terms that satisfy your reasonable-efforts duty, plus supervision. |
| Client facts in free/consumer ChatGPT | No | Consumer terms allow training on inputs. This is the Rule 1.6 breach scenario. |
| Citing AI research without checking it | No | The sanctions cases you've read about. Competence and candor duties: every citation gets verified. |
Why this matters more for small firms
Because the hours are already impossible. Clio's Legal Trends Report data puts average billable time at 2.6 hours of an 8-hour day, and Thomson Reuters' State of U.S. Small Law Firms found small-firm lawyers spend only about 60% of their time practicing law; the rest goes to running the firm. AI genuinely can return some of those hours (intake workflows, first drafts, scheduling, follow-up). The trap is grabbing the free consumer tool to do it and stepping on Rule 1.6 in the process.
The practical rule for your office
One sentence, in writing, for everyone including you: "Nothing about a client goes into any AI tool unless the firm has approved that tool in writing." Then approve tools deliberately: ones with contractual zero-retention and no-training terms.
Related: The same rules for medical and dental practices (HIPAA) · What can AI actually do for a small business?
Sources: ABA Formal Opinion 512 (2024); Illinois Supreme Court Commission (2Civility): ChatGPT ethics analysis; billable-hours data: Clio Legal Trends Report statistics; non-legal workload: Thomson Reuters: State of U.S. Small Law Firms; consumer-terms analysis: Spellbook: is it legal for lawyers to use ChatGPT.
Launching something new? We help owners plan the launch, build what it needs and get it seen, with software and AI handling the busywork.