AI Agents for Lawyers

Questions

Can Lawyers Use ChatGPT for Legal Work? Yes — With These Conditions

Lawyers can use ChatGPT for legal work on a non-training plan if they verify every citation, tell clients and bill honestly. The rules as of September 2026.

By Dr. Niklas SchmidtUpdated 13 November 20266 min read
On this page
  1. What the regulators actually say
  2. Condition one: the plan decides confidentiality
  3. Condition two: verify every citation yourself
  4. Conditions three and four: tell the client, bill the time
  5. A five-step setup that satisfies all four conditions
  6. Where this leaves you

Yes. As of September 2026 no bar in the United States, England and Wales, Germany or Austria prohibits lawyers from using ChatGPT, and the ABA, SRA, CCBE and ÖRAK all assume you will. Four conditions apply: use a plan that does not train on your inputs, verify every citation yourself, disclose where the rules or your client require it, and bill only for time actually spent.

What the regulators actually say

Every major bar has answered the question, and none of the answers is no. The ABA's Formal Opinion 512 (29 July 2024) applied six existing duties to generative AI and created no new rule; 17 US jurisdictions plus the ABA had formal AI opinions by June 2026. The English, European and Austrian positions are more recent and more pointed.

BodyPositionWhat it means for ChatGPT
ABA Formal Opinion 512 (29 Jul 2024)Competence, confidentiality, communication, candour, supervision, reasonable feesPermitted; you answer for the output; informed consent where a self-learning tool may expose client data
SRA Warning Notice (17 Aug 2026)42 misuse reports in a year; understand the tool "well enough to review its output critically"A consumer account you cannot explain to a client fails the test
CCBE Guide (2 Oct 2025)Inform clients and let them objectAI use belongs in the engagement letter
ÖRAK practice rules (Mar 2026)EU servers; disclose; lawyer keeps final responsibility; no client data for trainingConsumer plans fail the first and fourth rules
BRAK Hinweise (Dec 2024)Tools that store or forward client data for training breach § 43a Abs. 2 BRAOThe same conclusion in Germany

The detail sits on the ABA Opinion 512, SRA warning notice and ÖRAK, BRAK and CCBE pages. The short version: the tool is allowed; the lawyer is responsible.

Condition one: the plan decides confidentiality

ChatGPT is several products under one name. On Free, Go, Plus and Pro, OpenAI trains on your conversations unless you opt out. On Business ($25 per seat monthly, $20 annual), Enterprise, Edu and the API it does not train by default and offers zero data retention by contract (OpenAI enterprise privacy). GPT-6 Astra, public to paid users since 4 September 2026, stays switched off in Enterprise until an administrator enables it (9to5Mac): the firm, not the associate, decides.

ChatGPT planTrains on inputs by defaultFit for client material
Free, Go, Plus, ProYes, unless opted outNo. Public law, learning and your own non-client drafts only
BusinessNoYes, with a data processing agreement
Enterprise, APINo; zero data retention by contractYes, including privileged and regulated material under ZDR

Three courts have shown what happens on the wrong plan. In United States v. Heppner (S.D.N.Y., opinion 17 February 2026) Judge Rakoff held that research memos a defendant generated in consumer Claude were never privileged, because "Claude is not an attorney" and the terms permit logging and disclosure (Harvard Law Review). In UK v SSHD [2026] UKUT 81 (IAC) the Upper Tribunal said pasting client letters into public AI tools puts them "in the public domain". And on 5 January 2026 Judge Stein affirmed an order producing 20 million de-identified ChatGPT logs in the OpenAI copyright litigation (National Law Review); chat logs are discoverable. The privacy settings walkthrough shows the switches; the privilege question covers the case law.

Condition two: verify every citation yourself

Damien Charlotin's database listed 2,041 court decisions involving AI hallucinations on 14 September 2026, 811 of them involving lawyers rather than litigants in person (AI Hallucination Cases). The pattern dates from Mata v. Avianca (S.D.N.Y. 2023), where counsel asked ChatGPT to confirm its own cases; the price has risen. Couvrette v. Wisnovsky (D. Or., 12 December 2025) cost $110,204.38, most of it for refiling quietly instead of confessing (ABA Journal). At a hearing on 21 August 2026 in a New Mexico murder appeal whose brief contained ChatGPT-invented witness testimony, Justice C. Shannon Bacon asked counsel: "Do you watch the news? Do you listen to the radio?" (Reuters).

The benchmarks explain why. In HAQQ's 3,000-answer legal test (5 June 2026), 24% of answers across all models "cited or misapplied law that didn't support claims" (HAQQ). California's 2026 guidance states the rule: "No paper filed in court should contain citations that the attorney responsible for the filing has not personally read and verified" (Wilson Elser). The citation-checking protocol is the working version of that sentence.

Conditions three and four: tell the client, bill the time

Clients care more than lawyers assume. In Clio's 2025 Legal Trends Report, 78% of clients wanted to be told about AI use and 35% of lawyers rarely disclosed it. The CCBE and ÖRAK both say disclose. Illinois requires no disclosure in pleadings, but over 30 of the 94 US federal district courts had AI disclosure or certification requirements by April 2026. Check the client disclosure rules before the engagement letter goes out.

On fees, Texas Opinion 705 (February 2025) says you cannot charge for time "saved" by AI (Texas Center for Legal Ethics); Virginia LEO 1901 (24 November 2025) says a reasonable value-based fee need not shrink because AI was faster; California treats general AI subscriptions as overhead and bars mark-ups without informed written consent. The billing page reconciles them: bill hours actually worked, or agree a fixed fee up front.

A five-step setup that satisfies all four conditions

  1. Choose the plan. Business for a small firm; Enterprise with zero data retention for anything privileged or regulated. The personal Plus account stays personal.
  2. Fix the settings and record them. Switch off training and memory, take a dated screenshot, file it. That one act serves the SRA's "understand the tool" test and the EU AI Act's Article 4 training record.
  3. Write a one-page use rule. What may go in (public law, anonymised facts, your own drafts) and what never goes in on a consumer tier (identifiable client data, privileged strategy).
  4. Verify before it leaves your desk. Every citation opened in the primary source; a verification log on the matter file.
  5. Disclose and bill. One sentence in the engagement letter; time actually spent on the timesheet.

The AI Lab for Lawyers course sets up exactly these controls live in ChatGPT, Claude and Gemini before any client work is attempted.

Where this leaves you

ChatGPT is a legitimate legal tool on the right plan and a liability on the wrong one, and the right plan costs five dollars a month more. Models will keep changing (GPT-5.5 retires from ChatGPT on 14 October 2026); the four conditions have not moved since 2024. The ChatGPT for lawyers review covers what the current models do well; the confidentiality checklist covers the questions to ask any vendor before you sign.

Frequently asked questions

Is it legal for lawyers to use ChatGPT?

Yes. As of September 2026 no bar in the United States, England and Wales, Germany or Austria prohibits it. ABA Formal Opinion 512, the SRA's 2026 warning notice, the CCBE guide and the Austrian bar's practice rules all permit generative AI provided the lawyer protects confidentiality, verifies the output, keeps responsibility and, in Europe, tells the client. What gets lawyers sanctioned is filing what the tool invents.

Can I put client documents into ChatGPT?

Only on a plan that does not train on your inputs and under a data processing agreement, which means Business, Enterprise or the API rather than Free, Go, Plus or Pro. On consumer plans OpenAI trains by default, a federal judge has held that consumer-chatbot research memos were never privileged, and the UK Upper Tribunal has said pasting client letters into public tools puts them in the public domain.

Do I have to tell my client I used ChatGPT?

In Europe, effectively yes: the CCBE guide says inform clients and let them object, and the Austrian bar lists disclosure among its four practice rules. In the US no bar opinion requires blanket disclosure, but ABA Opinion 512 requires it where confidentiality or fees are affected, and 78% of clients say they want to know. One sentence in the engagement letter settles it.

Can I bill clients for time spent using ChatGPT?

You can bill the time you actually spend prompting, reviewing and correcting, and you can agree a fixed or value-based fee that reflects the result. You cannot bill hours the tool saved you; Texas Opinion 705 says so explicitly. California treats a general ChatGPT subscription as overhead that cannot be marked up or passed on without the client's informed written consent.

Put this into practice

AI Lab for Lawyers: a live, hands-on course for practising lawyers

4 live sessions (8 hours), recordings included, taught by Dr. Niklas Schmidt of Wolf Theiss. Rated 4.7/5 from 381 ratings; alumni from Norton Rose Fulbright, Paul Hastings, Pillsbury and Taylor Wessing.

Reserve a seat · €200 Maven money-back guarantee · certificate of completion

Tools

ChatGPT for Lawyers: GPT-6 Astra, ChatGPT Work and What to Switch Off

ChatGPT for lawyers as of September 2026. Which plans train on your chats, the 20-million-log discovery order, GPT-6 Astra, ChatGPT Work and a firm setup.

8 min read

Playbooks

AI Privacy Settings for Lawyers: Step-by-Step in ChatGPT, Claude, Gemini, Copilot and Perplexity

Exact steps to turn off training and check retention in ChatGPT, Claude, Gemini, Copilot and Perplexity as of September 2026, plus admin settings and a log.

10 min read

Questions

Does Using ChatGPT Waive Attorney–Client Privilege?

Using ChatGPT rarely waives privilege in the technical sense; the courts' answer is worse. In Heppner, consumer-chatbot memos were never privileged at all.

6 min read

Regulation & Ethics

ABA Formal Opinion 512 Explained: What US Lawyers Must Do When Using AI

ABA Formal Opinion 512 applies six Model Rules to generative AI: competence, confidentiality, communication, candour, supervision and fees. What each duty requires in 2026.

8 min read

Questions

Is ChatGPT GDPR-Compliant for Law Firms? A Practical Answer

ChatGPT can be used GDPR-compliantly by a law firm on Business, Enterprise or API terms with training off, a processor contract and a DPIA; consumer plans cannot. Seven steps.

7 min read

Guides

Which AI Tools Train on Your Data? A Confidentiality Checklist for Lawyers

Which AI tools train on your data, plan by plan as of September 2026: ChatGPT, Claude, Gemini, Copilot, legal vendors, plus bar rules and a 10-point checklist.

11 min read