06-23-2026
In a discovery motion before it, a Texas state judge reviewed ChatGPT “conversations” with the plaintiff’s principal, a non-lawyer, in Tate Group Automotive v. Legacy Automotive Capital. The plaintiff did not produce the conversations in discovery, claiming they qualified as attorney work product. The judge relied on Texas law that identifies work product as “material prepared or mental impressions developed in anticipation of litigation or for trial by or for a party” to reach his decision that the conversations may be protected. Work product can include materials generated by the principal of a corporate entity.
Two recent district court opinions, one in Michigan and another in Colorado, were cited by the Texas judge. Those courts reasoned that work product protection may be waived only when the materials are shared with an adversary or in a manner that disclosure to an adversary would be likely to occur. In Taft, the plaintiff did not share the materials either directly or in a questionable manner.
The Texas judge distinguished his decision from a New York federal court that reached the opposite conclusion. The New York court held that neither the attorney-client privilege nor the attorney work-product doctrine protected exchanges between a criminal defendant and a general AI platform. Inputting confidential information into a third-party platform, like ChatGPT, was tantamount to a voluntary disclosure outside the attorney-client relationship. According to the Texas judge, state rules “set forth a different standard for protectable attorney work product,” and it seemed to extend that protection to the Tate principal’s conversations with an AI tool.
However, the Texas judge did communicate limits to the protection, ordering Tate to hand over some discovery materials or products that it shared with AI, including those under a protective order. How far the protection extends, even under Texas law, is unclear.
Takeaway: As AI use grows, this area of the law will continue to evolve. Employers should not assume AI conversations are privileged and take precautions to protect sensitive and confidential information.