By Joshua Rubenstein, Louisiana Personal Injury Attorney | Blake Jones Law Firm, LLC | Licensed in Louisiana, Texas, Georgia, Washington and Alaska
Recognized by the National Trial Lawyers as a Top 10 Trucking Trial Lawyer, Joshua has spent nearly thirty years watching how a small piece of digital evidence can decide a case.
Everything you type into a search bar, post on social media, or ask an AI chatbot about your injury can end up in the defense’s file. I’ve watched good claims lose real value over one Facebook photo, and I’m now watching the same mistake happen with Google searches and ChatGPT conversations. Here’s what may be discoverable, what isn’t, and what to do instead.
Yes, when they’re relevant to your claim. Louisiana lets parties discover any non-privileged matter relevant to the case, or anything reasonably calculated to lead to admissible evidence. That standard sits in La. Code Civ. Proc. art. 1422. Federal courts hearing Louisiana cases apply a similarly broad relevance test under Fed. R. Civ. P. 26(b)(1), though the federal rule also weighs whether the discovery is proportional to the needs of the case.
Relevance decides everything.
A search for “best po-boy in New Orleans” or “best boudin in New Iberia” will likely never see the inside of a courtroom. A search for “how to injure yourself in a public place” almost certainly will. InfoTrack, a legal discovery technology firm, walks through that exact test and concludes a search like that would be “undoubtedly relevant, discoverable, and admissible” in a premises case.
In cases worth a combined $500 million and more, I have never once seen a personal search history help a client. I have seen it hurt—bigtime—and it hurts worse once you remember that Louisiana already reduces your recovery under comparative fault if a jury decides you share the blame. A bad search only pushes that number up—and if a jury finds you to be 51% or more at fault, you won’t recover anything. La. Civ. Code art. 2323, as amended by Acts 2025, No. 15 (eff. Jan. 1, 2026).
Insurance adjusters and defense firms monitor claimants’ public profiles routinely, which is why a good practice is always to limit viewership of your posts. That much you probably already suspect. What surprises people is how far behind a privacy wall discovery can reach.
In Nucci v. Target Corp., 162 So. 3d 146 (Fla. Dist. Ct. App. 2015), a Florida appeals court upheld an order compelling a slip-and-fall plaintiff to produce photographs from a Facebook account that held more than 1,200 images, set to “friends only,” limited to the two years before and after her accident. The Florida Bar News reported that the court found her privacy interest outweighed by the relevance of photos documenting her pre- and post-accident life. Friends-only is not lawyer-only.
Louisiana courts are moving in the same direction on a properly framed request for production. Setting your account to private slows the other side down. It does not stop them.
This is the newest problem, and most clients haven’t thought about it at all.
Talking to a chatbot feels private, almost like thinking out loud. It is not. Every prompt you type and every answer you get back lives on a company’s servers, and that record can get pulled into a lawsuit that has nothing to do with you.
In January 2026, a federal judge in the Southern District of New York ordered OpenAI to produce 20 million de-identified ChatGPT conversation logs in a copyright lawsuit, over the company’s own privacy objections, according to The National Law Review. That case has nothing to do with injury claims. It shows exactly what happens to chat logs once a court decides they matter to somebody’s dispute.
Now think about your own case. If you typed “how much is my back injury worth” or “what should I say to the adjuster” into ChatGPT, Claude, Perplexity, Gemini, Grok, or any other chatbot, that conversation is NOT privileged. Only communications with your actual lawyer carry that protection, backed by decades of Louisiana law your chatbot has never read. The bottom line: if you would not want it read aloud at a deposition, do not type it into a chatbot, a search bar, or a post.
In February 2026, a federal judge in the Southern District of New York put a name and a docket number on exactly this problem. In United States v. Heppner, No. 25-cr-00503-JSR (S.D.N.Y. Feb. 17, 2026), Judge Jed S. Rakoff ruled that 31 documents a criminal defendant created by querying Anthropic’s Claude chatbot were not protected by attorney-client privilege or the work-product doctrine, even after the defendant handed them to his own lawyers.
Bradley Heppner was under federal investigation for securities and wire fraud. Before agents ever seized his devices, he had been feeding Claude details about his case and asking it to organize his thoughts and outline defense arguments. He wasn’t a lawyer talking to a client. He was a defendant talking to a chatbot about his own case.
Judge Rakoff gave three reasons the privilege didn’t survive, as summarized in Gibson Dunn’s analysis of the ruling. Claude is not an attorney, so nothing exchanged with it counts as a communication between client and attorney. Heppner had no reasonable expectation of confidentiality, because the platform’s own terms let the company log, train on, and disclose his data. And because his lawyers never directed him to use Claude in the first place, the work-product doctrine, which protects material prepared in anticipation of litigation at counsel’s direction, never attached.
Sharing the documents with his lawyers afterward didn’t fix any of it.
Privilege does not attach retroactively.
Heppner is a criminal fraud case, not a civil Jones Act or trucking personal injury case, and I want to be careful not to overstate how far it reaches. But the reasoning travels. If you ask an AI chatbot to organize your thoughts about your accident, summarize what a witness told you, or draft talking points for your deposition, you’ve done the same thing Heppner did. Handing that conversation to me afterward doesn’t make it privileged. It never was.
Here’s the mistake that turns an embarrassing search into a real legal problem: deleting it.
While Louisiana law doesn’t allow a suit for negligently destroying evidence, Reynolds v. Bordelon, 2014-C-2362 (La. 6/30/15), 172 So. 3d 607, Louisiana does permit discovery sanctions on parties who intentionally destroy evidence. This means that the judge could tell the jury to infer that whatever was deleted would be harmful for your claim. In federal court, Fed. R. Civ. P. 37(e) allows sanctions specifically for destroying electronically stored information, including search history, chat logs, and social media posts.
Judges do not like being lied to, and juries like it even less.
If you already searched something you regret or asked a chatbot something you wish you hadn’t, the fix is simple: leave it alone and tell me. As we explain in our post on how much you can recover after a Louisiana car accident, the value of your claim rests heavily on credibility. I would rather explain an awkward Google search to a jury than explain a deleted one.
If you walked into my office tomorrow with an active claim, here’s the list I would run through with you:
None of this means going dark. It means keeping the parts of your life that matter to your claim inside a room with an attorney-client privilege around it, not scattered across social media, search engine and AI servers you don’t control.
Only the parts relevant to your claim, and usually only through the discovery process rather than a casual request. Searches tied to your injuries, your symptoms, or how to deal with an adjuster are the ones most likely to be found relevant and ordered produced.
No. Attorney-client privilege protects communications with your actual attorney. It does not extend to a chatbot, even one that sounds like it understands your situation. Those conversations sit on a company’s servers and can be swept into litigation, as a federal court showed when it ordered OpenAI to produce 20 million ChatGPT logs in January 2026.
Don’t delete it. Call me and tell me exactly what you wrote or asked. I can usually work with an awkward post. I cannot work with one a judge thinks you tried to hide.
Often, yes. Courts have ordered production of photos from accounts set to friends-only when the images are relevant to injury or damages claims, as in Nucci v. Target Corp. Privacy settings limit who sees your content voluntarily. They do not override a court order.
No, and a federal court addressed this exact scenario in February 2026 in United States v. Heppner. Handing your chatbot notes to your attorney after the fact does not make them privileged. The conversation with the chatbot was never protected to begin with, and sharing it later doesn’t reach back in time to fix that.
For more than fifty years, on the land, on the water, and on the roadways of America, our firm has fought for injured people. If you have questions about what you’ve posted, searched, or typed into an AI chatbot since your accident, call before you touch any of it.
Hancock Whitney Center
701 Poydras St # 4100,
New Orleans, LA 70139
Fields Marked With An "*" Are Required
"*" indicates required fields
Copyright ©2026 Blake Jones Law Firm, LLC • All Rights Reserved.
Digital Marketing By: 
*Images are obtained under license from Canva and other third-party stock image providers, with attribution included where required.
