Your Chatbot Is Not Your Lawyer: When AI Conversations Become Exhibit A

You typed the question at midnight. The answer felt private.

Maybe the prompt asked how to transfer money before a divorce. Maybe it described a disputed contract, an employee departure, an undisclosed account, or what to say in a demand letter. No lawyer was on the chat. No one else was in the room. Then the dispute became a lawsuit, and the other side asked for the conversation.

Can a ChatGPT conversation become evidence? Potentially. But the answer is not as simple as “AI chats are discoverable” or “private chats are protected.” Five separate questions matter: relevance, possession or control, privilege, preservation, and admissibility.

Private Is Not the Same as Privileged

People often use “private” and “privileged” as if they mean the same thing. They do not. Privacy describes an expectation about who will see information. Privilege is a legal protection that can prevent otherwise relevant communications from being disclosed.

Florida’s lawyer-client privilege protects qualifying confidential communications between a client and a lawyer made for the purpose of obtaining or providing legal services. A chatbot is not a lawyer. Asking an AI tool a legal question does not, by itself, create an attorney-client relationship or a blanket privilege over the conversation.

The analysis can become more complicated if counsel directed a client or litigation team to use a tool for a specific purpose, or if the chat includes or reflects counsel’s advice. Work-product and privilege questions are highly fact-specific. The safe assumption is not that every AI conversation is protected. It is that the issue should be reviewed before the material is produced, withheld, forwarded, or summarized.

When an AI Conversation May Matter

Discovery is not limited to documents that conclusively prove a claim. In Florida civil litigation, the inquiry generally turns on whether the material is relevant to a party’s claim or defense and proportional to the needs of the case. Florida family-law discovery incorporates the civil discovery framework, subject to the family rules.

An AI chat may become relevant when it records what a party was considering, what facts the party supplied, or what action the party planned. In a divorce, that could include prompts about transferring assets, valuing a business, explaining an unusual expense, changing account access, or timing a separation. In commercial litigation, it could include prompts about terminating a contract, taking customer information, responding to a preservation demand, calculating damages, or rewriting communications after a dispute arose.

Relevance still has limits. A demand for every AI conversation a person has ever had may be overbroad, intrusive, and disproportionate. A targeted request tied to a disputed transaction, date range, account, or issue is different. The wording, scope, burden, and available alternatives all matter.

Where the Evidence May Come From

The first source may be the user, not the AI company. Chats may exist in an account history, data export, browser history, downloaded file, screenshot, forwarded email, shared link, work device, synced phone, or backup. A conversation may also appear indirectly in a document that copied, paraphrased, or relied on it.

Third-party process presents a separate set of questions. A subpoena must be valid, properly issued, and directed through the appropriate procedure. Provider policies, jurisdiction, account identification, retention, burden, privacy interests, and objections may affect what can be obtained. The existence of a civil request process does not mean every request will succeed or that every historical chat remains available.

Deletion Can Create a Bigger Problem

Once litigation is filed or reasonably anticipated, relevant electronically stored information may need to be preserved. Deleting a chat, turning on auto-delete, clearing a device, or editing a shared record after that duty arises can create a dispute about lost evidence. The legal consequences depend on the duty to preserve, the steps taken, whether the information can be restored or replaced, and the resulting prejudice.

Do not try to “clean up” an account because a conversation is embarrassing or incomplete. Preserve first. Let counsel determine relevance, responsiveness, privilege, and whether a protective order or narrower production is appropriate.

A Screenshot Is Not the Whole Conversation

Even if a chat is produced in discovery, it does not automatically become trial evidence. Florida law requires authentication or identification. A cropped screenshot may omit the date, account, earlier prompts, later corrections, attachments, model-generated errors, or the surrounding exchange. The source file, export, device, metadata, and witness testimony may become important.

The distinction between user input and AI output is critical. A user’s prompt may reflect what the person wrote, believed, suspected, invented, or tested. The model’s response is generated content, not independent proof that the prompt was accurate. Admissibility may also raise hearsay, completeness, relevance, unfair-prejudice, and best-evidence questions.

What to Do When a Dispute Is Emerging

If a divorce or business dispute is becoming likely, identify which AI tools and accounts were used, whether they are personal or company-controlled, and whether chats are stored, shared, exported, or set to disappear. Preserve potentially relevant conversations in a defensible way. Do not access another person’s account without authorization, guess passwords, alter devices, or circulate sensitive material unnecessarily.

Counsel should also consider the opposite direction: whether an opponent’s AI use is genuinely tied to a disputed issue and whether a focused request can seek the relevant record without turning discovery into a search through someone’s entire digital life.

The Bottom Line

AI chats sit at the intersection of modern evidence and old legal rules. The technology is new. The questions are familiar: Who created the record? Who controls it? Is it relevant? Is it protected? Was it preserved? Can it be authenticated? And what does it actually prove?

The fact that a conversation felt private does not automatically keep it out of litigation. The fact that it was produced does not automatically make it admissible. The most important step is often the earliest one: preserve the record, stop improvising, and obtain advice before a private chat becomes the most misunderstood exhibit in the case.

MB Law Group represents individuals, business owners, professionals, and companies in Florida and New York family-law and commercial disputes involving financial records, digital evidence, contracts, businesses, and high-stakes litigation. Contact MB Law Group to discuss the record before it becomes the strategy.

Attorney Advertising. This article is for general informational purposes only and is not legal advice. Reading it does not create an attorney-client relationship. Prior results do not guarantee a similar outcome. Every matter depends on its facts and governing law.

Frequently Asked Questions

Can ChatGPT conversations be subpoenaed in a Florida divorce?

Potentially. A targeted request may seek relevant AI conversations within a party’s possession, custody, or control. Scope, proportionality, privilege, privacy, retention, and objections must be evaluated case by case.

Are questions I ask ChatGPT protected by attorney-client privilege?

Not automatically. A direct conversation with a chatbot does not create an attorney-client relationship. A separate privilege or work-product issue may arise in unusual circumstances involving counsel, but it requires a fact-specific legal analysis.

Can AI chats be used in a business lawsuit?

Potentially. Prompts or responses may relate to contract performance, intent, timelines, customer information, asset transfers, damages, or preservation. Relevance and discoverability do not by themselves establish admissibility or truth.

Should I delete embarrassing AI chats before filing for divorce?

No. If litigation is filed or reasonably anticipated, deleting potentially relevant information may create a preservation or spoliation issue. Preserve the material and obtain legal advice before changing account or retention settings.

Is a screenshot of an AI chat enough for court?

Not necessarily. The proponent may need to authenticate the record and establish completeness and context. Account data, exports, metadata, the device, testimony, and the distinction between user input and model output may matter.

Can the other side demand every AI conversation I have ever had?

A request that broad may draw objections based on relevance, proportionality, burden, privacy, or privilege. A focused request tied to a particular claim, transaction, person, date range, or account is more likely to present a serious discovery issue.

Does deleting a ChatGPT chat guarantee it is gone?

No assumption should be made about availability or retention. Copies may exist in exports, screenshots, shared links, emails, devices, backups, workplace systems, or material already preserved under a legal obligation.