Texas judge rules ChatGPT chats can be protected work product

The core of the dispute

A Texas Business Court recently addressed a novel legal question regarding the privacy of artificial intelligence interactions. The case, Tate Group Automotive LLC v. Legacy Automotive Capital LLC, involved a 2025 dealership acquisition lawsuit. During the discovery process, one of the defendants produced a massive amount of data involving their use of ChatGPT.

The defendant, Kris Tate, used the AI tool to assist with several high-level tasks. These included reviewing documents, performing complex analysis, and developing specific litigation strategies. Out of approximately 1,800 pages of ChatGPT conversations, the defendant attempted to withhold eight specific conversations from the opposing side.

The opposing party argued that these chats should be public because they were shared with a third-party technology vendor. They claimed that because the vendor had no duty of confidentiality, the defendant had effectively waived any legal protections. They relied on a 2026 federal case to argue that AI chats do not count as protected work product.

How the court reached a ruling

Judge Dorfman conducted an in camera review, which is a private examination of materials by the judge to determine their contents. He ultimately disagreed with the argument that using an AI tool automatically strips away legal protections. The judge compared AI tools to other types of litigation-support technologies already used by the legal profession.

The court found that the specific chats in question qualified for protection because they reflected the mental impressions of the defendant. Because these prompts and outputs were created in anticipation of litigation, they met the criteria for work product. The judge noted that the Texas Rules of Civil Procedure seem to extend this protection to the chats in question.

However, the judge did not grant total immunity for every single interaction. He ordered the production of dozens of other pages that did not meet the specific legal standard for protection. This distinction shows that while AI chats can be private, they are not automatically shielded from the discovery process.

State vs. federal standards

A significant part of the ruling hinged on the difference between state and federal legal standards. The defendants pointed to federal cases to argue that sharing information with a vendor without a confidentiality agreement waives privilege. The judge acknowledged these federal precedents but noted that Texas rules are different.

The judge observed that the state's rules plainly appear to extend protection to the materials in question. This suggests that the legal environment for AI usage may vary significantly depending on where a lawsuit is filed. It highlights a growing complexity for legal teams who must navigate different rules for the same technology.

The court also emphasized that protections are only waived if the materials are disclosed to an adversary. Sharing data with a technology platform is not the same as handing that data directly to the person on the other side of the courtroom. This distinction is crucial for how companies choose to integrate AI into their daily workflows.

Risks and future requirements

Despite the ruling, the court warned that using AI still carries significant confidentiality risks. Judge Dorfman left open the question of whether uploading sensitive discovery materials into ChatGPT violated existing protective orders. He suggested that litigants must be extremely careful about the data-control settings they use on these platforms.

The court recommended that the parties involved update their protective orders to be more specific. These documents should make it "unquestionably clear" how confidential information can be shared with third-party vendors. This move aims to prevent ambiguity when a company uses a tool like ChatGPT to process private data.

Moving forward, the court noted that this is a question of first impression nationwide. Because the case authorities cited by both sides were published in 2026, the law is still being written in real-time. Practitioners are encouraged to be proactive by discussing AI use with clients, vendors, and experts before a matter reaches the litigation stage.