Think Twice Before Asking ChatGPT Your Legal Questions: What Every Client Needs to Know About AI, Privilege, and Confidentiality

AI tools like ChatGPT, Microsoft Copilot, Claude, and Google Gemini are everywhere—and they can feel like having a lawyer in your pocket. Type in a question, get what looks like legal advice in seconds. But here is the critical point: talking to an AI chatbot is nothing like talking to your lawyer. Your conversations with AI are not confidential, not legally protected, and not always accurate. This alert explains why that matters and what you can do about it.

In this alert, we cover:

• Why your AI conversations have zero legal protection
• How sharing sensitive information with AI can come back to haunt you
• When AI is fine to use—and when you absolutely need a real lawyer
• Simple steps to protect yourself if you do use AI tools

What Is Attorney-Client Privilege (and Why Should You Care)?

The Short Version

Attorney-client privilege is one of the most powerful legal protections you have. In plain terms, it means that when you talk to your lawyer in confidence to get legal advice, no one can force either of you to reveal what was said—not the other side in a lawsuit, not the government, not anyone. This protection has existed for centuries and is considered fundamental to our legal system. 

The privilege encourages you to be completely honest with your lawyer—even about facts that might be embarrassing or damaging—because you know those conversations are protected. For the privilege to apply, a few conditions must be met:

1. You are seeking legal advice (not business or personal guidance)
2. From a licensed attorney acting as your lawyer
3. Your conversation relates to getting that legal advice
4. The conversation is kept confidential (not shared with outsiders)
5. You are the one communicating with the lawyer (or your lawyer is responding to you)
6. The protection lasts permanently
7. Neither you nor your lawyer can be forced to disclose the communication
8. Unless you voluntarily waive (give up) the protection 

Why this matters to you: Privilege protects not just your lawyer’s advice, but also the information you share with your lawyer so they can advise you properly. Think of it as a two-way shield: it protects what you tell your lawyer and what your lawyer tells you.

Important Limits: What Privilege Does NOT Cover

Privilege is powerful, but it has limits. It does not cover every conversation you have with someone who happens to have a law degree. Here are some common situations where privilege does not apply:

• Lawyers wearing a different hat: If you hire an attorney to act as a business consultant, financial advisor, or in some other non-legal role, those conversations are not privileged
• Documents you already had: If you had documents that could have been subpoenaed before you gave them to your lawyer, handing them to your lawyer does not suddenly make them privileged
• Crime or fraud: Privilege never protects communications made to help plan or carry out illegal activity 

The Big Risk: AI Is Not Your Lawyer, and Your AI Conversations Are Not Protected

ChatGPT Is a Tech Company—Not Your Attorney

This is the single most important point in this alert: When you type a question into ChatGPT, Copilot, or any AI chatbot, you are nottalking to a lawyer. You are sending information to a technology company. No attorney-client relationship exists. No privilege attaches. No confidentiality obligation protects you.

It does not matter that the AI’s response looks and sounds like legal advice. Because AI tools are not lawyers, nothing you tell them is protected by attorney-client privilege. Period.

Worse: Using AI Can Destroy Privilege You Would Otherwise Have

Here is where it gets really dangerous: if you share information with an AI platform before discussing it with your attorney, you may have permanently waived (given up) the privilege that would otherwise protect that information—even in your later conversations with your actual lawyer.

The legal rule is straightforward: If you voluntarily share privileged information with any third party, the privilege is waived. An AI chatbot is a third party. Using one is a voluntary disclosure. The fact that it is a computer rather than a person does not change the analysis. 

Even worse, waiving privilege on one communication can open the door to waiving privilege on all communications about the same subject. In other words, one careless AI query could expose an entire body of otherwise-protected attorney-client communications.

A Real-World Scenario: What Could Go Wrong

Imagine you type details about a legal dispute, sensitive business deal, or personal legal matter into ChatGPT. Here is what can happen:

1. No privilege protects that disclosure. As far as the law is concerned, you just told a stranger your secrets.
2. The company behind the AI has your data. Depending on the platform's terms of service, your input may be stored indefinitely, reviewed by employees, or used to train future AI models.
3. Your opponents can potentially get those records. In litigation or a government investigation, the other side may be able to subpoena your AI interaction history from the platform.
4. Your conversations with your actual lawyer may lose protection too. If you discussed the same subject with AI, you may have destroyed the confidentiality needed to keep your attorney communications privileged.

Your Data Is Not Safe: What AI Companies Do With Your Information

Where Does Your Information Go?

When you type something into an AI chatbot, it does not just disappear after you close the browser. Most platforms collect and keep what you share. Here is what that means in practice:

• Your conversations are not private. The company running the AI platform can access what you typed—even if no one else can see it.
• Your input may be used to train AI. Some platforms feed your conversations into their systems to make the AI smarter—which means your confidential information could become baked into the AI's knowledge base.
• Real people may read your conversations. AI companies sometimes have employees review user interactions for quality control purposes.
• Courts and regulators can demand your AI records. If you are ever involved in a lawsuit, investigation, or regulatory inquiry, the other side can subpoena your AI conversation history directly from the platform.

Our practical advice: Before you type anything into an AI tool, ask yourself: “Would I be comfortable if this showed up in a courtroom, in a newspaper, or on my opponent’s desk?” If the answer is no, do not type it.

AI Companies Have No Legal Duty to Keep Your Secrets

When you hire a lawyer, that lawyer is legally and ethically required to keep your information confidential. Violating that duty can result in disbarment and personal liability. 

AI platforms have no such obligation. They are not licensed professionals. They are not subject to bar discipline. They can change their privacy policies at any time. If your information leaks, you have no professional conduct complaint to file and likely no practical remedy.

AI Gets It Wrong—A Lot: The “Hallucination” Problem

AI Confidently Makes Things Up

Here is something most people do not realize: AI tools routinely make up information that sounds completely real. In the legal world, this includes fabricating court cases that never happened, inventing statutes that do not exist, and citing legal rules that are flat-out wrong. Experts call this “hallucination,” and it happens far more often than you might expect. Malkeet LNU v. Blanche, No. 24-4790, 2026 U.S. App. LEXIS 16174 (9th Cir. June 3, 2026).

The numbers are sobering:

• Even legal-specific AI tools (not just general chatbots, but products built specifically for lawyers) got answers wrong 17% to 33% of the time in 2024 testing. Malkeet LNU v. Blanche, No. 24-4790, 2026 U.S. App. LEXIS 16174 (9th Cir. June 3, 2026).
• Common errors include misrepresenting what a court actually decided, confusing one party’s argument with the court’s ruling, and getting the legal hierarchy wrong (for example, treating a trial court opinion as binding appellate law). Malkeet LNU v. Blanche, No. 24-4790, 2026 U.S. App. LEXIS 16174 (9th Cir. June 3, 2026).
• Multiple courts have already sanctioned (fined or punished) lawyers who submitted AI-generated legal research to courts without checking whether it was real. Malkeet LNU v. Blanche, No. 24-4790, 2026 U.S. App. LEXIS 16174 (9th Cir. June 3, 2026), Park v. Kim, 91 F.4th 610 (2d Cir. 2024).

What this means for you: If trained lawyers cannot trust AI output without verification, you certainly should not rely on it for decisions that affect your legal rights, your money, or your freedom. A confident-sounding AI response is not the same as actual legal authority. 

Even Lawyers Cannot Trust AI Without Double-Checking

Courts now hold that lawyers who use AI must independently verify every citation, every factual claim, and every legal conclusion the AI generates. Malkeet LNU v. Blanche, No. 24-4790, 2026 U.S. App. LEXIS 16174 (9th Cir. June 3, 2026). Simply prompting an AI, checking that citations look real, and calling it done is not enough—a lawyer must actually read the cases, reason through the law, and apply professional judgment. Malkeet LNU v. Blanche, No. 24-4790, 2026 U.S. App. LEXIS 16174 (9th Cir. June 3, 2026), United States v. Farris, 171 F.4th 920 (6th Cir. 2026).

The lesson for you is clear: if licensed attorneys with years of legal training are not allowed to take AI at face value, then you—without that training—should be even more cautious. You may not have the background to spot when an AI has invented a case, misread a statute, or given you advice that does not apply in your state.

When AI Is Fine—and When You Need a Real Lawyer

When AI Can Be a Useful Starting Point

AI tools can be genuinely helpful for certain limited purposes, as long as you do not share confidential information and do not treat the output as reliable legal advice:

• Learning the basics: Understanding general legal concepts, looking up what common legal terms mean, or getting a big-picture overview of how an area of law works
• Getting oriented: Figuring out what questions to ask your lawyer, or understanding generally what area of law might be relevant to your situation (while knowing the AI’s answer might be incomplete or wrong)
• Low-stakes, non-confidential tasks: Brainstorming language for a public blog post, organizing non-sensitive information, or drafting routine business correspondence that contains nothing confidential

When You Absolutely Must Talk to a Lawyer Instead

Put down the chatbot and pick up the phone to call your attorney when:

1. Your situation is unique. AI gives generic answers based on patterns in its training data. It cannot account for the specific facts of your case, your state’s particular laws, or the nuances that could make or break your position.
2. Confidential information is involved. If the matter touches trade secrets, personal financial details, sensitive business deals, details of a legal dispute, or anything you would not want an adversary to see—keep it off AI platforms entirely.
3. The stakes are high. Potential lawsuits, serious financial consequences, criminal exposure, regulatory compliance, or your fundamental rights are not matters to entrust to a chatbot that makes things up a third of the time.
4. You need someone who is accountable to you. Your lawyer owes you legal duties of competence, confidentiality, loyalty, and honesty. If they fail, you have remedies. An AI platform owes you nothing beyond its terms of service.
5. You need your communications protected. If there is any chance your communications could be sought by an opposing party, a government investigator, or anyone else, you need the shield of attorney-client privilege—and that only exists when you talk to a real lawyer.
6. Documents will go to a court or government agency. Courts have already sanctioned lawyers for submitting AI-generated filings with fabricated citations. Malkeet LNU v. Blanche, No. 24-4790, 2026 U.S. App. LEXIS 16174 (9th Cir. June 3, 2026), Park v. Kim, 91 F.4th 610 (2d Cir. 2024). Any document heading to a court or agency needs to be prepared or reviewed by a competent attorney.
7. You are facing litigation, an investigation, or a legal claim. Active legal proceedings are high-stakes, time-sensitive, and subject to strict procedural rules. Do not use AI to draft legal responses, answer subpoenas, or handle any aspect of active litigation without your attorney’s guidance.

Five Rules to Protect Yourself If You Use AI

If you decide to use AI tools for the limited, appropriate purposes described above, follow these five rules to minimize your risk:

Rule 1: Strip Out All Identifying and Confidential Details

Before you type anything into an AI platform, remove:

• Remove all names (your name, your company's name, names of other individuals or entities involved)
• Remove dates, locations, and other identifying details
• Remove account numbers, financial data, and proprietary business information
• Remove any facts that are confidential, sensitive, or could harm you if disclosed

Example: Instead of asking, “Can ABC Corporation fire John Smith for refusing to work on Sundays because of his religion under California law?” try: “Can an employer in California terminate an employee for refusing to work on a particular day due to religious beliefs?” Same question—zero identifying details.

Rule 2: Assume Everything You Type Will Become Public

Before you hit “send,” imagine your AI query being read by:

• The lawyer on the other side of your dispute
• A government investigator or regulator
• A journalist writing a story about you or your company
• Your biggest business competitor

If any of those scenarios makes you uncomfortable, that is your answer: do not type it in.

Rule 3: Never Enter Legally Protected Information

Specific categories of information that should never go into an AI tool:

• Anything covered by a nondisclosure agreement (NDA)
• Trade secrets or proprietary business information
• Anything you have discussed with your lawyer (attorney-client privileged communications)
• Terms of any legal settlement
• Medical records, health information, or other regulated personal data
• Anything a court has ordered to be kept confidential

Rule 4: Never Trust AI Legal Information Without Verification

• Assume every legal citation is fake until proven otherwise. If an AI gives you a case name, statute number, or regulation, verify it independently before relying on it in any way.
• Do not make legal decisions based on AI output. Always confirm legal information with a licensed attorney or an authoritative legal resource before acting on it.
• Never use AI-generated content in court filings or official submissions without having a qualified attorney review and verify every word.

Rule 5: Always Consult Your Attorney Before Acting

Use AI for background research and general curiosity only. Before you make any legal decision, sign any document, respond to any legal communication, or take any action that could have legal consequences, talk to a licensed attorney. Your lawyer can give you advice tailored to your specific facts, your jurisdiction, and your goals—and that conversation will be protected by privilege.

Quick Reference: AI Tools vs. Your Attorney

The Bottom Line

AI chatbots are useful tools for general research and learning—but they are not lawyers, they are not confidential, and they are not reliable. When you type information into an AI platform, you are handing it to a third party with no legal obligation to protect it. That disclosure is not privileged, may be stored indefinitely, and could come back to haunt you in litigation or an investigation.

Before you open ChatGPT for a legal question, ask yourself three things:

• Does this involve confidential information or high stakes?
• If yes—call your attorney instead
• If you do use AI—strip out all identifying details, and never trust the output without independent verification

When in doubt, call me. The cost of a conversation with your attorney is a fraction of the cost of waiving privilege, exposing confidential information, or making decisions based on fabricated legal guidance.

This alert is provided for informational purposes only and does not constitute legal advice. For guidance specific to your situation, please contact us directly.