ChatGPT for Lawyers: 7 Use Cases, 4 Things to Never Paste

“Quick approval?” lands in your inbox on a late Friday afternoon, attached to a vendor MSA sales has already promised to close. You pull the indemnification and limitation-of-liability sections, strip the identifying details, and ask ChatGPT which terms fall outside market for a mid-market SaaS deal. What comes back is a serviceable issue list. Some of it matches what you would have flagged yourself, some of it is generic enough to skip, and the call on what your company will accept is still yours to make. That is ChatGPT for lawyers at its most useful, and getting there takes reps.

Michael A. Jacobs, a former Morrison and Foerster partner who litigated Oracle v. Google and Apple v. Samsung, told CZ and Friends, GC AI’s podcast for in-house legal leaders:

“The AI tool is a good second or third year associate right now. And so whatever you would do with a second or third year associate if you’re the partner is what you should do with the AI tool, except you have to be doubly worried about hallucinations.”

The second half of that is where lawyers get in trouble. A second-year who cites a case has read it. ChatGPT will hand you cases that do not exist, in the same tone it uses for the real ones.

GC AI is the enterprise-grade legal AI platform Cecilia Ziniti built for in-house teams after serving as general counsel three times, at Anki, Bloomtech, and Replit. As of July 2026 it is used by 1,900+ legal teams, including 200+ public companies and 25 unicorns. The roster includes Snyk, Liquid Death, Columbia Sportswear, and Hitachi.

GC AI runs that same first pass inside the Microsoft Word document you already have open. Citations come back as character-level verbatim text from your own files through Exact Quote, and research answers arrive with the primary-law source attached through Research.

Can Lawyers Use ChatGPT?

Yes, with duties attached. ABA Formal Opinion 512, issued July 2024 by the American Bar Association, confirmed that generative AI use by lawyers is permissible under existing ethics rules, so long as counsel maintains three obligations: competence, supervision, and confidentiality. California, New York, Florida, and the D.C. Bar have published guidance built on the same three pillars.

Those duties belong to the lawyer, and no platform can discharge them for you.

Competence means understanding what the model does well and where it fails, including hallucination patterns and jurisdictional blind spots. Supervision means treating output like a junior associate’s draft that gets checked before it gets signed. Confidentiality means knowing what happens to the text you paste, where it lives, who can read it, and whether the platform trains on it. The AI sanction cases that made headlines since 2023 trace back to a gap in one of those three.

What Lawyers Use ChatGPT For

ChatGPT does its best work on the repetitive, first-draft, and translation tasks that fill an in-house lawyer’s day. A competent human reviews every output before it leaves the building. These are the categories that hold up:

• Summarizing long documents. Compress a 30-page research memo into a three-bullet executive version, or a deposition transcript into the five facts that matter.

• First-draft client communications. Draft a plain-English email that explains a regulatory update to a non-lawyer stakeholder, then edit for tone and accuracy.

• Issue spotting as a second pass. Run a clause you have already reviewed through the model as a second set of eyes, catching the term you skimmed past.

• Outlining briefs. Build the section-by-section skeleton of a motion before you fill in the citations yourself.

• Converting dense statutory text. Turn GDPR Article 28 or DGCL Section 203 into a client-readable summary a business partner can follow.

• Getting up to speed in an unfamiliar jurisdiction. Orient yourself on procedural requirements or local rules before a matter lands, then verify against primary law.

• Prepping before you call outside counsel. Work the question far enough to know what you are asking, so the hourly call starts at the hard part.

GC AI customers run these same categories, with the supervision layer built into the workflow from the start.

ChatGPT Prompts for Lawyers

A structured prompt beats a one-line ask. Give the model a role, the context, the specific task, your constraints, and the output format. Here are four you can adapt today, drawn from the our AI prompts for lawyers in-house teams run daily.

Contract issue spotting (second pass):

You are my second set of eyes on a SaaS vendor agreement I have already
reviewed. I represent the customer. Read the indemnification and
limitation-of-liability sections below and list every term that falls
outside market standard for a mid-market SaaS deal, ranked by risk to the
customer. For each, give the clause, the risk in one sentence, and a
suggested redline.
[paste clauses]

You are my second set of eyes on a SaaS vendor agreement I have already
reviewed. I represent the customer. Read the indemnification and
limitation-of-liability sections below and list every term that falls
outside market standard for a mid-market SaaS deal, ranked by risk to the
customer. For each, give the clause, the risk in one sentence, and a
suggested redline.
[paste clauses]

You are my second set of eyes on a SaaS vendor agreement I have already
reviewed. I represent the customer. Read the indemnification and
limitation-of-liability sections below and list every term that falls
outside market standard for a mid-market SaaS deal, ranked by risk to the
customer. For each, give the clause, the risk in one sentence, and a
suggested redline.
[paste clauses]

You are my second set of eyes on a SaaS vendor agreement I have already
reviewed. I represent the customer. Read the indemnification and
limitation-of-liability sections below and list every term that falls
outside market standard for a mid-market SaaS deal, ranked by risk to the
customer. For each, give the clause, the risk in one sentence, and a
suggested redline.
[paste clauses]

Plain-English client email:

Statutory summary:

Brief outline with a hallucination guardrail:

You are helping a licensed attorney outline a motion to dismiss for failure
to state a claim in federal court. Based on the fact summary below, produce
a section-by-section outline with the legal standard, the three strongest
arguments in order, and the counterargument to pre-empt for each. Leave
placeholders for citations. Do not invent case citations.
[paste fact summary]

You are helping a licensed attorney outline a motion to dismiss for failure
to state a claim in federal court. Based on the fact summary below, produce
a section-by-section outline with the legal standard, the three strongest
arguments in order, and the counterargument to pre-empt for each. Leave
placeholders for citations. Do not invent case citations.
[paste fact summary]

You are helping a licensed attorney outline a motion to dismiss for failure
to state a claim in federal court. Based on the fact summary below, produce
a section-by-section outline with the legal standard, the three strongest
arguments in order, and the counterargument to pre-empt for each. Leave
placeholders for citations. Do not invent case citations.
[paste fact summary]

You are helping a licensed attorney outline a motion to dismiss for failure
to state a claim in federal court. Based on the fact summary below, produce
a section-by-section outline with the legal standard, the three strongest
arguments in order, and the counterargument to pre-empt for each. Leave
placeholders for citations. Do not invent case citations.
[paste fact summary]

That last line matters on every prompt that touches law: tell the model to leave citation placeholders so you supply verified cases yourself.

How to Prompt Well

The skill ceiling on legal prompting runs higher than most lawyers expect. The GC AI 101 class teaches it with a scene: channel your inner Meryl Streep, being exceedingly clear about expectations, while the AI is Anne Hathaway, smart and eager and waiting to understand the assignment. Context is king.

A reliable base prompt carries five parts:

1. Role. “You are drafting for a licensed attorney representing the landlord.”

2. Context. The lease, the facts, the jurisdiction, your position.

3. Task. One specific ask, not three stacked together.

4. Constraints. Tone, length, audience, what to avoid.

5. Format. Bullet list, two paragraphs, a table, a redline.

If building that structure every time is the friction, Easy Prompt turns a plain-language thought-starter into an optimized legal prompt, and the Skill Library saves the ones that work so the rest of your team can run them.

Specificity does the work. “Draft a two-paragraph summary of California AB 2013 for our engineering leads, neutral tone, no legalese” produces a usable first pass. “Tell me about AB 2013” produces a hedged one. The reflexive disclaimers are default system behavior, and the reasoning underneath handles substantive work once you give it a well-structured prompt. That gap between the first-week user and the six-month user closes with reps.

What You Should Never Paste Into ChatGPT

Four categories stay out of ChatGPT Free and Plus. Each maps to a real risk that a February 2026 ruling made concrete.

• Client-identifying data. Names, company identifiers, matter-specific facts, deal terms. Pasting these into a consumer tier is disclosure to a party outside the attorney-client relationship.

• Privileged material. Strategy memos, draft pleadings, internal legal analysis, anything you would claim as work product. If opposing counsel can subpoena the chat record, you have waived privilege by typing.

• Jurisdiction-specific analysis without verification. Niche state statutes, recent amendments, unsettled areas where the model’s training cutoff hides a 2025 change behind a confident answer.

• Final filings without a citation check. If Westlaw, LexisNexis, or Casetext cannot return the case, the case does not exist. Check every citation before you sign.

One SDNY ruling put that risk on the record. In United States v. Heppner (SDNY), Judge Jed S. Rakoff held on February 17, 2026, that a criminal defendant’s written exchanges with a generative AI platform were neither privileged nor protected work product.

Rakoff’s three-part test: the AI is not an attorney, so the communication is not attorney-client privileged; the platform’s Terms of Service permitted training on user inputs, which broke confidentiality; and the defendant used the tool on his own, outside counsel’s direction. The court left one door open under the Kovel doctrine, which extends privilege to agents retained by counsel: counsel-directed use of a platform with contractual confidentiality may preserve privilege. A consumer tool a client picks up alone satisfies none of the three prongs.

The sanction cases run on a parallel track. In [Mata v. Avianca](https://en.wikipedia.org/wiki/Mata_v._Avianca,_Inc.) (SDNY 2023), an attorney submitted a brief citing six cases ChatGPT fabricated, and Judge P. Kevin Castel sanctioned him $5,000. The failure mode repeats across Park v. Kim and Wadsworth v. Walmart: unverified citations relied on in a filing. Our breakdown of the Heppner ruling walks through the privilege math, is ChatGPT confidential covers what counts as a breach, and is ChatGPT private covers what OpenAI does with your inputs.

The Compliance Gap: ZDR, SOC 2, and What ChatGPT Enterprise Leaves Open

ChatGPT Free and ChatGPT Plus use prompts for model training by default as of July 2026, which puts any client data you paste into the training set and outside the attorney-client relationship. ChatGPT Enterprise and ChatGPT Business change the picture: both offer zero data retention options and contractual confidentiality, and OpenAI publishes SOC 2 Type 2 reports for the Enterprise product.

Comparison data as of July 2026.

The practical question is scope. What is certified, who issued the report, which sub-processors are covered, and whether the contract indemnifies a privilege breach or only a direct data leak. Buyers evaluating ChatGPT Enterprise for privileged work should confirm each of those in writing. Enterprise-grade security and fitness for privileged work are two different questions.

:Tricia Kinney, General Counsel at BlueLinx, put the underlying point on the CZ and Friends podcast

“I am a huge fan of using legal specific AI tools as opposed to consumer specific AI tools. You want them training in the same context that we’re operating in. Legal language is very different.”

GC AI’s posture maps to the three prongs by design. GC AI is SOC 2 Type II and SOC 3 certified, GDPR compliant, with zero data retention agreements with OpenAI and Anthropic, and AES-256 encryption. Each customer’s data lives in a logically isolated database, and GC AI does not train on customer data. The deployment is counsel-directed, the confidentiality is contractual, and the platform functions as an agent of the legal team.

What You Get From GC AI That ChatGPT Does Not

In-house teams ask what a purpose-built legal AI platform does that a general-purpose one cannot. In the May 2026 In-House Legal Bench, a 100-task study built on 1,200+ attorney-developed criteria and validated by LLM-as-judge against human review, the accuracy gap was clear:

• GC AI: 86.8%

• ChatGPT (GPT-5.5): 79.8%

• Claude (Opus 4.7): 68.4%

• Gemini (3.1 Pro): 57.5%

That gap compounds across hundreds of contracts and research queries. Three differences drive it.

Citations you can open. ChatGPT can generate citations that do not exist, the Mata v. Avianca failure mode built into any general-purpose model. GC AI surfaces verified sources through Exact Quote, which pulls character-level verbatim text from your uploaded documents, and through Research, which biases toward primary law and government sources and returns the link with the answer. Ritesh Patel at Viant Technology used to Google privacy questions or call outside counsel by the hour:

“It’s also replaced Googling. Now my first stop is GC AI. I describe the setup, get an answer with citations, and use that to brief my team or our business partners.”

Your company’s context. By default, ChatGPT does not know your MSA template, your redline defaults, or your prior matters. GC AI connects to all of it through Files, Projects, and Playbooks, so answers arrive anchored to how your team operates. Alexis Palmer, Senior Managing Counsel at Snyk, describes the difference:

“I can put a thought-starter into the platform and it gives me a lovely prompt back. With other tools, you need a perfectly polished prompt for it to work well. With GC AI, I don’t.”

Work inside Word. Redlining, issue spotting, and drafting happen in the document you already have open through GC AI for Word, with no browser tab to switch to. Laura Knight, VP of Legal at Secure Code Warrior, put it plainly:

“GC AI’s Word Add-in is in a class of its own compared to other legal AI tools I have evaluated.”

Trisha Mauer, VP of Legal at Tonal, describes the same pull:

“I go straight to GC AI for everything from research requests to litigation responses.”

Joys Choi, Senior Director of Legal at Tipalti, saved 609 hours in a single year, the equivalent of 76 working days.

Switch from ChatGPT to a purpose-built platform when your in-house team grows past three lawyers, when privileged work becomes routine, or when client data falls under a regulated framework. For the side-by-side, see the GC AI vs. ChatGPT comparison, and if you’re weighing Claude, read our GC AI vs. Claude comparison.

Frequently Asked Questions

Can Lawyers Use ChatGPT?

Yes. Under ABA Formal Opinion 512 (July 2024), lawyers may use generative AI if they maintain three duties: competence, supervision, and confidentiality. State bars including California, New York, Florida, and D.C. follow the same framework. The duties belong to the lawyer, and the AI sanction cases that made headlines since 2023 trace back to a gap in one of the three.

What Legal Tasks Can ChatGPT Handle?

ChatGPT works for summarizing research memos, drafting plain-English client communications, issue spotting in contracts as a second pass, outlining briefs, and converting dense statutory text into readable summaries. A competent lawyer reviews and verifies every output before it leaves the building. GC AI customers run these categories with built-in source verification through Exact Quote and Research.

What Are the Best ChatGPT Prompts for Lawyers?

The strongest legal prompts give the model a role, context, a specific task, constraints, and an output format, then instruct it to leave citation placeholders instead of inventing cases. Effective examples include contract issue spotting, plain-English client emails, and statutory summaries. GC AI publishes 15 tested prompts in-house teams use daily.

What Data Should Lawyers Never Paste Into ChatGPT?

Lawyers should keep client-identifying data, privileged strategy memos, draft pleadings, and matter-specific facts out of ChatGPT Free and Plus. Pasting these into a consumer tier is disclosure to a party outside the attorney-client relationship, which is the failure Judge Rakoff identified in United States v. Heppner. GC AI’s zero data retention architecture and logically isolated databases address this risk directly.

Does Attorney-Client Privilege Protect ChatGPT Conversations?

No. In United States v. Heppner (SDNY, February 2026), Judge Rakoff held that written exchanges with a consumer-grade generative AI platform are not privileged, because the AI is not an attorney, the Terms of Service permitted training on inputs, and the client used the platform on his own, outside counsel’s direction. Counsel-directed use of a platform with contractual confidentiality may preserve privilege under the Kovel doctrine.

Can ChatGPT Chats Be Subpoenaed or Used in Court?

Yes. Under United States v. Heppner (SDNY, February 2026), written exchanges with a consumer-grade AI platform are treated as disclosure to a third party, which waives both attorney-client privilege and work product protection. If the platform’s terms permit training on user inputs or sharing with third parties, those chats are discoverable. Counsel-directed use of a platform with contractual confidentiality is treated differently under the Kovel doctrine.

Does ChatGPT Enterprise Satisfy Post-Heppner Privilege Requirements?

ChatGPT Enterprise and ChatGPT Business offer zero data retention and contractual confidentiality that consumer tiers do not, making them materially safer for legal work. Whether Enterprise fully satisfies the Heppner test depends on specific contract terms, which sub-processors are covered, and whether deployment is genuinely counsel-directed. A platform designed around all three Rakoff prongs gives a cleaner answer.

Have Any Lawyers Been Sanctioned for Using ChatGPT?

Yes. Courts have sanctioned attorneys in Mata v. Avianca, Park v. Kim, and Wadsworth v. Walmart, with penalties including fines and referrals to grievance committees. In every case, the offense was filing a brief with AI-generated citations that were never verified against Westlaw, LexisNexis, or Casetext before signing.

What Should a Legal Team’s ChatGPT Policy Include?

A 2026 ChatGPT policy should address which tier is approved (Enterprise or Business with zero data retention is the floor for client-adjacent work), data-handling rules (no client-identifying data or draft pleadings in consumer tiers), citation verification (every AI citation checked against Westlaw, LexisNexis, or Casetext before signing), and disclosure obligations where a court or client contract requires them. GC AI’s counsel-directed architecture and zero data retention agreements address all four by design.

Which AI Is Best for Legal Work: ChatGPT, Claude, or Gemini?

GC AI scored highest in the May 2026 In-House Legal Bench, a 100-task study GC AI ran using 1,200+ attorney-developed criteria: GC AI 86.8%, ChatGPT (GPT-5.5) 79.8%, Claude (Opus 4.7) 68.4%, and Gemini (3.1 Pro) 57.5%. The study compared GC AI against those three general-purpose models. The largest gaps appeared in regulatory tracking, legal research, and checklists, where legal-specific context carries the most weight.

What Are the Best ChatGPT Alternatives for Lawyers?

The best ChatGPT alternatives for lawyers are purpose-built legal AI platforms that keep client data confidential and cite real sources. GC AI is built for in-house teams, with zero data retention agreements with OpenAI and Anthropic, SOC 2 Type II certification, and citations you can open and verify. For a full comparison of the options, see our guide to the best legal AI tools for in-house counsel.