In 2023, a lawyer filed a brief citing six cases to support his client. Every single one was fake. ChatGPT had generated them confidently, cleanly, and completely wrong.
That case put the entire legal profession on notice, as reported by Forbes.
ChatGPT is now inside law firms worldwide. It’s drafting documents, summarizing depositions, and cutting hours off routine work. It’s also creating real liability for lawyers who trust it too much.
Here you’ll find everything you need to use it correctly: the real use cases, the risks that have already ended careers, and the practices that keep you protected.
What Is ChatGPT and What Does It Actually Do in A Legal Context?
ChatGPT generates text by predicting the next word based on patterns in its training data. It doesn’t make sense. It doesn’t verify. It produces output that sounds authoritative.
In law, that distinction is everything.
Neither the free nor the paid version checks what it produces against verified legal databases. That responsibility is entirely yours every time, with every output.
Will ChatGPT Replace Lawyers?
No, but it will replace the parts of your practice that don’t require you.
Lawyers who lose ground won’t be replaced by AI. They’ll be outpaced by lawyers who use it smarter. According to the Clio Legal Trends Report, 70% of clients are neutral toward or prefer firms that use AI.
Can Lawyers Legally and Ethically Use ChatGPT?
Yes, but using it doesn’t transfer responsibility.
On July 29, 2024, the ABA Standing Committee on Ethics and Professional Responsibility released Formal Opinion 512, covering generative AI in legal practice. It confirmed that competency, confidentiality, client communication, candor, supervision, and fee rules all apply.
As of early 2026, 35+ state bar associations have issued their own AI guidance. The ABA hasn’t banned AI, but it requires competent, transparent use. You must understand the tool, verify all output, and disclose AI use when court rules require it.
Check your state bar’s guidance before building any AI-assisted workflow.
How Lawyers Are Using ChatGPT Right Now
ChatGPT isn’t replacing legal work. It’s handling the time-consuming parts without requiring judgment.
Document drafting: First drafts of NDAs, demand letters, engagement agreements, and legal briefs come together faster. You still review everything, but a structured draft beats a blank page. Legal brief writing is now a desired AI feature for 30% of legal professionals, according to the 2025 Legal Industry Report.
Document and deposition review: Upload a transcript or discovery file. ChatGPT pulls key facts, flags inconsistencies, and builds timelines faster than manual review. Medical record work that once took a paralegal 30–40 hours now takes under an hour with AI.
Legal research with hard limits: Use it to understand unfamiliar concepts and map potential arguments. Never use it to verify citations. That step requires Westlaw, LexisNexis, or Casetext, no exceptions.
Case strategy: Feed it your facts and ask it to stress-test your position before it matters in court. Identify weaknesses before opposing counsel does.
Litigation support: Motion outlines, cross-examination drafts, chronology summaries. ChatGPT handles the scaffolding so you can focus on the argument.
Client correspondence: Plain-language summaries and routine updates, drafted fast and reviewed before they leave your desk.
How Fast Is AI Adoption in Law Firms Right Now
The legal profession has moved faster than most expected.
AI adoption within the legal profession is nearly three times higher year over year, rising from 11% in 2023 to 30% in 2024, according to the ABA Legal Technology Survey Report. By late 2025, that momentum had accelerated further. The 8 am Legal AI Adoption Report found 69% of legal professionals personally using AI tools for work, up from 31% the prior year.
Here’s what the adoption curve looks like across three years:
- 2023: 11% of law firms using AI tools (ABA Legal Technology Survey)
- 2024: 30%, nearly three times higher in one year (ABA Legal Technology Survey)
- 2025: 79% of legal professionals use AI in some capacity (Clio Legal Trends Report)
- 2025–2026: 69% confirmed personal daily use, more than doubling in a single year (8am Legal AI Adoption Report)
Note: the 79% and 69% figures measure different things. The Clio figure covers any AI use across the profession. The 8am figure covers individual daily personal use. Both point in the same direction.
The economic case is hard to ignore. According to Thomson Reuters projects, AI could save lawyers four hours per week and create $100,000 in additional billable capacity per attorney annually
What Went Wrong in The Sanctions Cases You Need to Know
This is where theory meets consequence. The consequences have been severe and escalating.
Mata v. Avianca (S.D.N.Y., 2023)
Attorneys Steven Schwartz and Peter LoDuca filed a brief citing six cases that did not exist. ChatGPT fabricated every citation, docket number, judicial opinion, and quoted passage.
Judge Castel fined both attorneys $5,000 and required them to notify every judge whose name appeared in the fake opinions.
Park v. Kim (2nd Circuit, 2024)
Attorney Jae S. Lee filed an appellate brief containing nonexistent citations generated by ChatGPT.
The Second Circuit referred her to the grievance committee, the first time the pattern reached the appellate level. The referral is documented in the Second Circuit’s February 2024 order.
Wadsworth v. Walmart Inc. (D. Wyoming, February 2025)
Judge Kelly H. Rankin fined Morgan & Morgan attorney Rudwin Ayala $3,000 and revoked his pro hac vice admission. Supervising attorney T. Michael Morgan and local counsel Taly Goody were each fined $1,000.
Eight cases cited in the motions did not exist. Critically, the hallucinated citations came from a legal AI platform, not ChatGPT. That fact is documented in Judge Rankin’s February 2025 sanctions order.
November 2025 (Prosecutor misconduct)
A prosecutor submitted multiple briefs riddled with AI-generated errors, explaining she had resorted to AI due to working on multiple matters simultaneously and moving too fast. The Boston Bar Association noted the misuse could have resulted in wrongful convictions and unconstitutional sentencing.
Every sanctioned lawyer made the same mistake, signing a filing without opening Westlaw, LexisNexis, or Casetext to confirm the cited cases actually existed.
The Real Risks Every Lawyer Needs to Know
ChatGPT’s usefulness is real. So are the ways it damages careers when used carelessly.
- Fabricated citations: ChatGPT will cite cases that don’t exist, quote statutes never written, and do it with complete confidence. You won’t know unless you verify every reference in a trusted legal database.
- Privilege exposure: Client facts entered into a public AI model leave your control instantly. The American Bar Association’s Model Rule 1.6 makes clear that confidentiality obligations apply regardless of the tool used. Attorney-client privilege does not extend to OpenAI’s servers.
- PII in unencrypted systems: Names, financial records, and medical information belong nowhere near a public model. The FTC’s guidelines on data privacy make clear that liability for exposed client data falls on the professional who handled it.
- Billing ethics violations: ABA Formal Opinion 512 addresses fees directly. Billing a client for the full time a task would have taken manually when AI completed it in minutes may violate your duty to charge reasonable fees. Develop a billing policy for AI-assisted work before you need one.
- Over-reliance drift: The most dangerous failure isn’t one bad output. It’s gradually trusting the tool more than your own judgment until errors reach clients and courts. The 2025 prosecutor case documented by the Boston Bar Association is the clearest example of this pattern.
- Sanctions and malpractice: Courts have already sanctioned attorneys for unverified AI output. The Federal Rules of Civil Procedure Rule 11 requires attorneys to certify that filings are accurate. You signed it. You own it entirely.
Every risk on this list already has a paper trail attached to a real lawyer’s career.
Best Practices for Using ChatGPT Safely as A Lawyer
Using ChatGPT without a clear set of rules is how lawyers end up in disciplinary hearings.
- Verify every citation independently: Never trust a case reference ChatGPT produces. Check every citation against Westlaw, LexisNexis, or Casetext before it goes anywhere near a filing. This is the single step every sanctioned attorney skipped.
- Treat every output like a first draft: Review it the way you’d review a junior associate’s memo, carefully, critically, before anyone else sees it.
- Never input client details into public models: Names, case facts, PII none of it. Use anonymized placeholders instead.
- Build a mandatory review step: Under ABA Model Rules 5.1 and 5.3, any AI-assisted work product must be reviewed by a supervising lawyer before it leaves the firm, with the review logged.
- Check your state bar guidance: As of early 2026, 35+ state bars have issued AI-specific guidance. Some states go further than the ABA framework. Ignorance is not a defense.
ChatGPT vs. Purpose-Built Legal AI What’s Actually Different
General-purpose AI and legal-specific AI are not the same. The gap matters for your clients and your license.
| Feature | ChatGPT (public) | Legal AI (Harvey, Clio Duo, LexisNexis AI) |
|---|---|---|
| Citation verification | None | Built-in against verified databases |
| Client data privacy | No guarantees | Designed for attorney-client confidentiality |
| Case management integration | None | Direct integration with firm systems |
| Ethics guardrails | None | Built around Model Rules frameworks |
ChatGPT works for drafting and brainstorming. The moment client data, case citations, or court filings are involved, purpose-built legal platforms are the safer choice.
What Law Students Should Know About ChatGPT?
Law school is where habits form. The ones you build around AI now will follow you into every courtroom.
| Area | How ChatGPT helps | Where it can hurt |
|---|---|---|
| Case law study | Summarizes complex rulings in plain language | May misstate holdings or fabricate citations |
| Exam preparation | Outlines arguments and flags counterpoints | Can produce confident but incorrect analysis |
| Legal writing practice | Generates first drafts to edit and improve | Over-reliance stunts independent drafting ability |
| Academic integrity | Speeds up research and organization | Submitting AI output as your own violates most school policies |
| Bar exam readiness | Explains concepts across jurisdictions | Cannot replace memorization under exam conditions |
Note: Most law schools now actively flag AI-generated submissions. Know your institution’s policy before you draft anything.
Where This Is All Heading: AI in Legal Practice
The rules around AI in law are being written right now. What’s acceptable today may be required or prohibited within two years.
- Court disclosure requirements are spreading: Multiple federal courts now require attorneys to declare whether AI was used in drafting filings. The Northern District of Texas AI Standing Order was among the first. More courts are following.
- Large firms are building formal governance: Internal AI policies, mandatory training programs, and dedicated oversight roles are becoming standard. Individual judgment alone is no longer considered sufficient.
- Adoption is accelerating fast: A global Thomson Reuters survey found that legal organizations actively integrating generative AI jumped from 14% in 2024 to 26% in 2025.
- Regulation is running behind the technology: New AI tools are released faster than ethics boards can evaluate them. The gap between what exists and what’s been formally assessed is widening every month.
- Technological competence is now a legal duty: Under ABA Model Rule 1.1, staying current with technology is a core component of basic competence, not optional professional development.
Conclusion
ChatGPT is a capable tool in the right hands. In the wrong ones, it’s a $5,000 fine, a grievance referral, and a story that runs nationally.
It fails due to fabricated citations, exposure of privilege, and overreliance that quietly compounds until it reaches a courtroom.
The lawyers who came looking for a straight answer on whether to use ChatGPT now have one. Use it. But use it with a verification step you never skip and a review process you never shortcut.
Start by checking your state bar’s AI guidance before you build any AI-assisted workflow into your practice.
Frequently asked questions
Can ChatGPT give legal advice?
No. It explains legal concepts but cannot account for your jurisdiction, your facts, or your professional obligations. It is a drafting tool, not a legal opinion.
What happened to the lawyer who cited fake cases from ChatGPT?
New York attorney Steven Schwartz was fined $5,000 after filing a brief with six fabricated citations. Judge Castel also required him to notify every judge named in the fake opinions.
Is ChatGPT reliable for legal research?
No. It produces plausible-sounding output but cannot verify citations. Every reference it generates must be checked against Westlaw, LexisNexis, or Casetext before use.
