Anything you type into an AI chatbot can be stored, retained, and later handed over in response to a subpoena or discovery request. It is not a private diary, and it is not a confidential conversation with your lawyer. Recent reporting has documented a fast-growing trend: conversations people have with tools like ChatGPT are increasingly being subpoenaed and introduced as evidence in both civil lawsuits and criminal investigations. For anyone going through a divorce, a custody dispute, a protection order case, or a criminal charge in King County, that trend carries real risk — and most people have no idea it exists.
At our firm, we are already advising clients to treat AI chatbots the same way they should treat a text message or an email: as something that can be read aloud in a courtroom someday. Here is what you need to understand before you type.
What "Discoverable" Actually Means
When you use a consumer AI chatbot, your messages do not stay on your device. Every prompt is transmitted to the provider's servers, where it is stored along with metadata — the account it came from, the date and time, and often a device or IP address. Courts have spent two decades treating emails, texts, and cloud files as fair game in litigation, and AI chats are simply the newest category of the same thing.
In a Washington civil case, the rules of discovery are broad. Under Washington Superior Court Civil Rule 26, a party may generally obtain any non-privileged material that is relevant to the claims or defenses in the case, including electronically stored information. A spouse's attorney in a divorce, or the other parent's attorney in a custody fight, can serve a discovery request or a subpoena aimed squarely at your AI chat history. In a criminal investigation, law enforcement can seek the same records from a provider through a search warrant or subpoena. The provider — not you — decides how to respond, and providers routinely produce records when legally compelled.
There Is No "AI Privilege"
This is the point people find most surprising. The law recognizes a handful of privileged relationships — attorney and client, physician and patient, and in Washington, spouses and certain counselors — that keep specific communications out of court. Your conversation with a chatbot fits none of them.
The head of the company that makes ChatGPT, Sam Altman, has said so publicly. He has warned that when people use ChatGPT as a therapist or confidant, there is no legal confidentiality protecting what they share, and he has called for lawmakers to create some form of "AI privilege" that does not currently exist. Until that changes — and it may never — assume that a chatbot conversation carries no more legal protection than a public social media post.
That matters because of how people talk to chatbots. Users tend to be strikingly candid, typing things they would never say out loud: venting about a spouse, describing an argument, asking how to move money, or narrating an incident in detail. Those same words, offered up voluntarily and stored on a company's servers, can become a roadmap for the other side.
Deleting the Conversation Does Not Delete the Record
Many people assume that clearing their chat history erases it. It usually does not. Providers retain data on their own servers according to their own retention policies, independent of what you see in the app.
This is not hypothetical. In the ongoing federal litigation The New York Times Co. v. Microsoft Corp. and OpenAI (S.D.N.Y.), a court ordered OpenAI to preserve output log data that would otherwise have been deleted — including users' deleted and temporary chats — while the case proceeds. In other words, a court order can require a provider to keep the very conversations you thought you had thrown away. Once a lawsuit is filed or reasonably anticipated, the duty to preserve relevant evidence attaches, and "I deleted it" is not a defense — it can instead expose a party to sanctions for destroying evidence.
How This Plays Out in a Washington Family Law or Criminal Case
The abstract risk becomes concrete quickly in the kinds of matters we handle:
- Divorce and property division. Asking a chatbot how to shield an asset, downplay income, or time a large withdrawal can surface later as evidence of an intent to dissipate marital property — an issue a court weighs directly in dividing the estate under RCW 26.09.080.
- Child custody and parenting plans. Messages that reveal anger, substance use, or a plan to relocate can be used to challenge your judgment or your account of events in a parenting dispute.
- Domestic violence and protection orders. A candid, blow-by-blow description of an incident typed into a chatbot can be introduced to contradict your testimony — or the other party's.
- Criminal defense. Search queries and chatbot conversations about the very conduct under investigation are among the first things prosecutors look for, and they can speak directly to knowledge and intent.
- Title IX and school discipline. Written accounts drafted or "workshopped" with an AI tool can be obtained and compared against later statements for inconsistencies.
In each of these settings, the danger is the same: words you meant as private thinking-out-loud become a prior statement that a skilled opposing attorney can use against you.
What You Should Do
You do not have to swear off technology. You do need to use it with the assumption that nothing you type is truly private:
- Do not discuss your case with an AI chatbot. Not the facts, not your strategy, not your feelings about the other party. Treat it like a conversation held in a crowded room.
- Do not use a chatbot to draft statements, declarations, or messages about your dispute. If you need something written, work with your attorney, whose communications are privileged.
- Assume deletion does nothing. Do not rely on clearing your history, and never destroy anything once a case is filed or anticipated.
- Keep confidential matters inside privileged channels. The one conversation that the law protects is the one you have with your own lawyer.
Frequently Asked Questions
Can the other side really get my ChatGPT history in a divorce? Potentially, yes. Washington's civil discovery rules reach relevant, non-privileged electronically stored information, and a chatbot history is not privileged. The other party can request it from you directly or serve a subpoena on the provider. Whether a particular request succeeds depends on relevance and the judge, but the door is open.
What if I delete the conversation? Deleting it from the app usually does not remove it from the provider's servers, and a court can order the provider to preserve exactly those deleted chats — as one federal court has already done. Worse, deleting records after a case is filed or reasonably anticipated can itself be treated as destroying evidence and expose you to sanctions.
Is talking to an AI chatbot like talking to my lawyer or a therapist? No. Attorney-client and counselor communications can be privileged; a chatbot conversation is not, no matter how personal it feels. Only the conversation you have with your own attorney is protected.
Are private or "incognito" AI modes safe? Do not count on it. Retention varies by provider and service tier, and a court order can override a provider's ordinary deletion practices. Treat every mode as if it is being recorded.
Is it ever safe to use AI during my case? For neutral, non-case tasks unrelated to your dispute, the risk is low. The rule of thumb is simple: never put the facts of your case, your strategy, or your feelings about the other party into any AI tool.
Talk to a Seattle Attorney Before You Type
If you are facing a divorce, a custody matter, a protection order, or a criminal charge, the safest assumption is that your digital life — including your AI chatbot history — may be examined. The attorneys at Blair Kim Moeller, PLLC help clients throughout Seattle and King County protect themselves, and that increasingly means understanding where the new technology creates old-fashioned evidence problems.
Call us at 888-478-4001 or schedule a consultation through our website. A short conversation now can keep a careless keystroke from becoming a problem later.
Reviewed by [Reviewing Attorney], Blair Kim Moeller, PLLC. Last updated September 3, 2026.
The information on this website is for general information purposes only. Nothing on this site should be taken as legal advice for any individual case or situation. This information is not intended to create, and receipt or viewing does not constitute, an attorney-client relationship.
