When Clients Come in with AI Answers
At Nuddleman Law Firm, we increasingly see clients walk in the door after first consulting an AI chatbot—ChatGPT, Copilot, Gemini, Claude, and similar tools—about their legal problems. They often bring printouts or screenshots and say, “This is what the AI told me. Is that right?”
Sometimes the AI gets part of the story correct, but more often the analysis is incomplete, oversimplified, or just wrong. Even more troubling, using AI this way can create problems that go far beyond bad information: it can jeopardize attorney‑client privilege, expose confidential information, and complicate your legal position before a lawyer ever sees the file.
Losing the Attorney-Client Privilege
What Privilege Is—and Why It Matters
The Attorney‑client privilege is one of the strongest protections the legal system offers. When you talk to a licensed attorney in confidence to get legal advice, those communications are protected from disclosure to the other side, to courts, and to most government agencies. The protection encourages candid conversations: you can tell your lawyer the good, the bad, and the ugly, knowing those discussions are shielded.
Privilege requires four things: you are seeking legal advice, from a lawyer acting in that role, about a legal matter, and the communication is kept confidential—not shared with outsiders. When those conditions are met, neither you nor your lawyer can be forced to reveal what you discussed, unless you voluntarily waive the privilege.
How AI Use Can Waive Privilege
Here is the key problem: AI chatbots are third parties. They are not your lawyer, and they are not part of your privileged communication. If you take confidential facts, strategy, settlement terms, or your lawyer’s advice and type them into an AI tool, you are voluntarily disclosing privileged
information to a third party. Under longstanding privilege rules, that disclosure can destroy the protection you otherwise would have had.
The risk exists both before and after you talk to an attorney:
- Before you hire a lawyer. If you describe a sensitive dispute, internal investigation, or potential claim to a chatbot, and later hire counsel, the other side may argue that you waived privilege by first sharing those facts with a third‑party technology company.
- After you speak with your lawyer. If you summarize what your attorney told you, upload privileged emails, or ask AI to “check” your lawyer’s advice, you may be exposing privileged communications and undermining the confidentiality the privilege requires.[1]
Courts have also recognized that waiver can extend beyond a single conversation. When you waive privilege on one communication about a subject, it can open the door to discovery of communications on that same topic. That means a single ill‑considered AI query about a sensitive issue can put a whole body of otherwise‑protected attorney‑client communications at risk.[1]
Privilege vs. “Free Legal Advice”
It may be tempting to think of AI as “free legal advice” when you are worried about costs. But privilege is part of what you are buying when you hire a lawyer, and AI does not give you legal advice (yet!). Your attorney is ethically bound to keep your information confidential and to protect privileged communications; a chatbot is not.
The cost of trying to “save money” with AI first can be far higher if it leads to waiver of privilege on critical issues. Once privilege is gone, it is hard—often impossible—to get it back. That is why we recommend talking to a lawyer before sharing sensitive legal facts with any AI system.[1]
AI Tools Are Not Private or Protected
Privacy Policies Are Not Ironclad Contracts
Many AI platforms reassure users with privacy policies and marketing claims about confidentiality. But privacy policies are not negotiated attorney‑client agreements; they are typically one-sided statements a company can revise. Courts and commentators have repeatedly noted that these policies rarely create firm contractual rights for users and rarely prevent companies from sharing data in ways permitted by their terms or required by law.
Once you click “I agree” and start typing, your data enters the company’s systems. You do not control how long it is stored, how it is processed, or what secondary uses may be made of it—such as model training, internal analytics, or data sharing with affiliates and service providers.[5][1]
What Really Happens to Your Data
Most major AI providers log user interactions, both to improve their models and to monitor for misuse. That means your “private” legal question may be:
- Stored on company servers for an extended period.
- Reviewed by employees or contractors for quality‑control or safety checks.
- Used as training data to refine future versions of the model.
In litigation or regulatory investigations, opposing parties and agencies can sometimes seek those records directly from the AI provider. If the company receives a subpoena, court order, or law‑enforcement request, it may have to turn over user data despite earlier privacy assurances. If the company suffers a data breach, your supposedly private legal questions and the facts you reveal may be exposed in a hack.
Real‑world experience with privacy policies shows that users often believe these documents guarantee strict confidentiality, yet the policies themselves typically allow broad internal use and sharing—and rarely prevent disclosures compelled by law or exposed through security incidents. Once legal issue is in an AI system, you cannot reliably control where that information goes next.
For clients, the practical rule is simple: if you would be uncomfortable seeing your AI chat in a courtroom exhibit, a government file, a notification letter, do not put those facts into the chatbot.
AI Is Good, But Not Necessarily Correct
How Large Language Models Actually Work
Modern AI chatbots are built on large language models (LLMs). Technically, these systems are trained to predict the next word—or “token”—in a sequence, based on patterns learned from huge datasets. They are great at generating text that looks coherent and authoritative, but they do not “understand” the law the way a trained attorney does, and they do not independently verify the accuracy of their statements.
When you ask a legal question, the model is not consulting a court’s docket, Shepardizing ® cases, or checking statutory history. It is combining patterns from its training data and your prompt into a plausible answer. If your prompt is incomplete, one‑sided, or based on incorrect assumptions, the model will build on that foundation, often without challenging it.
Hallucinations and Legal Error
Researchers who have systematically tested LLMs on legal tasks have found high rates of hallucination: fabricated case citations, misstatements of holdings, and legal rules that either do not exist or apply to a different jurisdiction or period. Even specialized legal AI tools have shown significant error rates, and general‑purpose chatbots perform worse when asked detailed, verifiable legal questions.
Courts have sanctioned lawyers who submitted briefs containing AI‑generated, non‑existent cases, underscoring that “looks like a case citation” differs from real precedent. If experienced attorneys cannot take AI legal output at face value, individuals and businesses should be even more cautious about relying on it for high‑stakes decisions.
The Problem of One-Sided Facts
AI’s accuracy also depends heavily on the facts you choose to share. Most people present their situation in a way that highlights favorable details and reduces unfavorable ones. That is human nature—but it can skew AI’s output. Because the model cannot interview witnesses, review documents, or test your story against other evidence, it assumes the facts you provide are complete and correct.
A good attorney does the opposite. At Nuddleman Law Firm, we:
- Ask follow‑up questions to uncover missing details.
- Explore how the other side is likely to view the same events.
- Consider alternative legal theories, defenses, and procedural options.
We draw on experience and training to identify legally significant facts you may not realize matter and to correct misunderstandings about how the law applies in your specific jurisdiction and context. That nuanced, two‑sided analysis is something an LLM cannot do (yet) when it is limited to whatever you type into a chat box.
A Better Way to Use AI—and When to Call Us
AI tools can still play a limited, useful role. They can help you learn basic legal vocabulary, get a general overview of an area of law, point you in the right direction for where to research or find more information, or organize non‑confidential information. But they should not be the primary source of advice for decisions that affect your rights, your business, or your livelihood.
Before you plug a legal problem into a chatbot, ask yourself:
- Am I about to share facts I want to keep confidential?
- Could this issue lead to litigation, government investigation, or significant financial consequences?
- Would I be comfortable if my AI conversation were read by the other side or by a judge?
If the answer to these questions are “no,” it is time to talk to a lawyer instead of a bot.
Talk to Nuddleman Law Firm Before You Ask AI
If you are dealing with a workplace dispute, employment claim, or HR compliance issue in California, Nuddleman Law Firm is here to help. We understand both the promise and the pitfalls of AI, and we can advise you that is:
- Tailored to your facts and your goals.
- Grounded in actual law and procedure.
- Protected by attorney‑client privilege and professional confidentiality.
Before you rely on a chatbot, waive privilege, or expose sensitive information online, contact Nuddleman Law Firm to discuss your situation directly with an experienced employment attorney.