Texas just made the AI privilege fight about proof, not panic
A Texas Business Court minute entry did not bless ChatGPT as magic. It treated the chat log like work product, which leaves lawyers with the old questions: what was shared, who saw it, and whether anyone preserved the record.
An AI-assisted editorial, reviewed by a human before publishing. It reasons over our own tracker data (and cited context). A point of view, not legal advice.
The familiar scene has not changed much: a lawyer opens a file, a client swears something stayed confidential, and the other side wants to know whether the machine saw it too. What changed is that a Texas Business Court minute entry now puts that question in the open. According to the court’s minute entry, a represented plaintiff’s ChatGPT conversations could qualify as work product, and using ChatGPT did not automatically waive that protection.
That is not a blessing for chatbot secrecy. It is the same old work-product fight wearing a newer coat, with the added nuisance that the coat may have been left in a shared folder.
Texas did not bless ChatGPT as magic. It treated the chat log like potentially protected work product, and left the burden where it always was.
The old lesson was never about novelty
Lawyers already know the shape of this problem from the sanctions orders. In Mata v. Avianca, Inc., the court fined Schwartz and co-counsel $5,000 after a brief cited six ChatGPT-fabricated cases. In Whiting v. City of Athens, Tennessee, the Sixth Circuit sanctioned two Tennessee lawyers for appellate briefs with more than two dozen fake or misrepresented citations and facts. In Shahid v. Esaam, the Georgia Court of Appeals struck a divorce order that relied on bogus cases and sanctioned attorney Diana Lynch.
The point of those orders was never that generative AI is forbidden. The point was that filing something you have not checked is still filing something you have not checked. The machine does not take the blame when the brief lands.
Texas is pushing the issue into the proof phase
The Texas Business Court minute entry, as reported, relied on two 2026 federal cases and flagged a separate problem: sharing confidential material with AI can still create serious evidentiary and protective-order issues. That matters because it shifts the fight away from a crude yes-or-no question about AI use and toward the parts lawyers actually lose on: what was disclosed, how it was preserved, and whether the record can support waiver or protection.
That is where the real discipline lives. A party that wants work-product protection will need to prove the boundaries of the ChatGPT exchange, not just announce that the tool was used for thinking. A party arguing waiver will look for the opposite: disclosure, sloppiness, and the kind of factual gap that turns a claimed privilege into a story the court does not have to believe.
The market has moved faster than the habits
That tension is not confined to one Texas dispute. LawNext reported that the 2026 Report on the State of the US Legal Market found technology spending in firms grew 9.7 percent and knowledge-management spending 10.5 percent in 2025, while billable hours rose 2.5 percent. Thomson Reuters has argued that AI will force changes in billing and says each lawyer could save 190 work-hours a year, with 80 percent of law firm survey respondents expecting AI to have the biggest impact on the industry over the next five years. Clio says 79 percent of legal professionals are already using AI in their practice, while 70 percent of clients are agnostic or prefer firms that use it.
That is the part some lawyers still miss. The market is not pausing for anyone to work out whether AI feels proper. The market is already using it, pricing it, and asking whether the result is reliable.
Verification is now the baseline duty
That is why the minute entry lands at the same moment as the sanctions wave. Our own tracking shows 28 AI-related sanctions. Recent federal decisions, as reported, make clear that lawyers may use generative AI, but they must verify every citation and quote before filing. The duty is not abstract. It is mechanical, and courts are treating the failure as a professional lapse, not a software glitch.
If a lawyer wants a concrete model for the posture courts are rewarding, it is not mystical faith in the tool. It is line-by-line checking, with the source tethered to the claim. Turbo Law, which publishes this site, says its platform hyperlinks and line-cites assertions back to the source document. That is not a substitute for judgment. It is the minimum answer to a court that now expects proof rather than promises.
The bigger lesson is narrower than the hype
Texas did not declare ChatGPT privileged. It did something more practical and more useful: it treated the chat log like potentially protected lawyer work product, while leaving room for waiver, disclosure, and protective-order trouble. That is the sort of ruling that matters to a working lawyer because it tells you exactly where the danger sits.
The danger is not that AI exists in the office. The danger is that too many lawyers will treat a machine-generated draft as if the burden of checking somehow moved with the text. The court did not move that burden. It left it where it has always been, sitting on the lawyer’s desk, waiting for a signature.