Artificial intelligence tools such as ChatGPT, Claude, Gemini, and similar platforms are now part of everyday life. People use them to draft emails, summarize documents, organize thoughts, and understand complicated topics. In litigation, however, using public AI tools without guidance can create serious legal risks.
The main issue is not that AI is “bad.” The issue is that litigation involves confidential facts, attorney-client communications, legal strategy, settlement positions, and evidence. When those materials are entered into a public AI tool, the user may be disclosing sensitive information to a third-party platform. That disclosure can create problems for attorney-client privilege, attorney work product, discovery, confidentiality, and case strategy.
AI Is Not Your Lawyer
A public AI chatbot is not an attorney. It does not represent you. It does not owe you the same professional duties that a lawyer owes a client. It may also be subject to its own terms of service, data retention policies, logging practices, subpoenas, court orders, and other legal obligations.
This matters because attorney-client privilege generally protects confidential communications between a client and attorney for the purpose of seeking or providing legal advice. The privilege is powerful, but it can be waived. One common way to waive privilege is by sharing privileged information with a third party.
That means a client should not assume that a conversation with a public AI tool is private, privileged, or protected from discovery.
The Emerging Court Split
Recent federal cases have started to address whether AI prompts, AI outputs, and AI-assisted litigation materials are privileged or protected work product.
In one widely discussed federal case, United States v. Heppner, the court held that the defendant’s communications with a public AI tool were not protected by attorney-client privilege or the work-product doctrine. The court focused on the fact that the AI tool was not the defendant’s lawyer, the communications were not made at counsel’s direction, and the defendant had disclosed information to a third-party platform.
Other courts have taken a more protective approach in different circumstances. For example, in Warner v. Gilbarco, a federal court found that a self-represented litigant’s use of ChatGPT to help prepare litigation materials could be protected by the work-product doctrine. That result does not mean every AI chat is protected. It shows that courts are looking closely at the facts: who used the tool, why it was used, what was entered, whether the use was connected to litigation preparation, and whether confidentiality was preserved.
The safest lesson is this: AI use in litigation is not automatically privileged, and it is not automatically discoverable in every situation. But it is risky enough that clients should not use public AI tools for case-related matters without first speaking with their attorney.
Five AI Mistakes Clients Should Avoid During Litigation
1. Pasting Attorney Advice Into AI
A client should not paste an attorney’s email, legal analysis, litigation strategy, settlement recommendation, or case evaluation into a public AI tool to “get a second opinion” or “make it easier to understand.”
Even if the client has good intentions, this can disclose privileged legal advice to a third party. Once privilege is waived, it may be impossible to fix.
2. Uploading Case Documents for Summary or Analysis
Many people use AI to summarize contracts, bank records, text messages, court filings, deposition transcripts, photographs, or discovery responses. In an ordinary non-legal setting, that may seem convenient. In litigation, it can be dangerous.
Case documents may contain confidential information, private personal data, trade secrets, settlement communications, privileged material, or attorney work product. Uploading those documents to a public AI platform may create a discoverable record and may compromise confidentiality.
3. Drafting Messages to the Opposing Side With Too Much Context
AI can make an email sound calm and professional. But the risk is usually in the prompt, not just the final email.
For example, a client might paste a long explanation of the dispute, emotional frustrations, private facts, and proposed strategy, then ask AI to draft a polite response to the opposing party. The final email may look fine, but the prompt may contain damaging admissions or confidential strategy.
In litigation, what is typed into the AI tool may be more dangerous than what the AI produces.
4. Venting About the Case to a Chatbot
Litigation is stressful. Clients may feel tempted to use AI as a private place to vent, especially late at night. That can be risky.
Statements such as “I know I should not have said that,” “I just want to make them suffer,” or “I am afraid the documents look bad for me” may become harmful if they are later discovered or used for impeachment. Even emotional, informal, or exaggerated statements can create problems in court.
5. Assuming Deleted AI Chats Are Gone Forever
Deleting a chat from an app does not necessarily mean all associated data immediately disappears from every server, backup, log, or legal hold system. If litigation is pending or reasonably anticipated, deleting potentially relevant records may also create spoliation concerns.
The better practice is not to enter sensitive case information into a public AI tool in the first place.
Can Attorneys Use AI?
Yes, attorneys may use AI in appropriate ways. But attorneys must use professional judgment.
An attorney may use AI to help with general drafting, organization, brainstorming, editing, translation, tone, issue spotting, or formatting. Attorney-directed AI use may also fall within attorney work product when it is used to prepare for litigation. But that protection is strongest when the attorney controls the process, avoids unnecessary disclosure, uses appropriate confidentiality safeguards, and independently reviews the final work.
AI should be treated like a tool, not a lawyer. It can assist with drafting, but it should not replace legal judgment, factual verification, legal research, or attorney supervision.
Safer Ways to Use AI in Legal Matters
For clients, the safest rule is simple:
Do not put case facts, attorney advice, legal strategy, settlement discussions, private documents, text messages, financial records, or emotional commentary about your case into a public AI tool unless your attorney specifically tells you it is safe to do so.
For attorneys and law firms, safer practices may include:
- Using AI only under attorney supervision.
- Avoiding client-identifying facts unless necessary.
- Redacting names, dates, account numbers, private facts, and litigation strategy.
- Using enterprise or legal-specific AI tools with stronger privacy, retention, and confidentiality controls.
- Turning off training on user data where possible.
- Reviewing the AI tool’s terms of service, privacy policy, retention policy, and data-sharing practices.
- Maintaining an internal AI-use policy.
- Adding an AI-use warning to client intake materials.
- Asking new clients whether they have already used AI in connection with the dispute.
Suggested Client Instruction
Clients involved in litigation should be told clearly:
“Please do not use ChatGPT, Claude, Gemini, or any other public AI tool to discuss this case, summarize documents, draft messages, analyze evidence, or interpret attorney advice unless our office approves it first. AI use may create discoverable records and may affect attorney-client privilege or work-product protections.”
This instruction is not meant to scare clients away from technology. It is meant to prevent avoidable mistakes.
Bottom Line
AI can be useful in litigation, but it must be used carefully. The biggest risk is not that AI writes poorly. The bigger risk is that sensitive information may be disclosed to a third-party platform and later become part of the case.
Clients should not treat public AI tools as private legal assistants. Attorneys can use AI more safely when they control the workflow, protect confidentiality, and apply independent legal judgment.
In litigation, every communication matters. AI prompts and uploads are no exception.