Via legaltech.news
Rulings in Texas and New York protect ChatGPT conversations as attorney work product, but a Connecticut decision shows the shield has limits
In back-to-back decisions on June 3 and June 4, 2026, courts in Texas and New York ruled that AI-generated prompts and outputs created during litigation preparation qualify as protected work product. The decisions mark the first significant judicial endorsements of the idea that using AI tools doesn’t automatically open your legal strategy to opposing counsel.
Two states, two days, one message #
The Texas Business Court got there first. In Tate Group Automotive, LLC v. Legacy Automotive Capital, LLC, the court ruled on June 3 that ChatGPT conversations generated by a party principal are safeguarded as protected work product. The key reasoning: engaging an AI tool doesn’t inherently waive the protections that would apply if an attorney had done the same analysis with a legal pad and a highlighter.
One day later, the New York Supreme Court reached a strikingly similar conclusion in Assini v. Hayward. The court quashed subpoenas that targeted a litigant’s AI prompts and outputs, grounding its decision in work-product protections under New York’s Civil Practice Law and Rules, commonly known as CPLR.
The Connecticut counterpoint #
Before anyone declares total victory for AI privacy in litigation, there’s a wrinkle. Just weeks earlier, on May 18, 2026, US Magistrate Judge Thomas O. Farrish in Connecticut went the other direction entirely.
In Conservation Law Foundation, Inc. v. Shell Oil Co., Judge Farrish ordered the production of AI prompts used by an expert witness. The reasoning was different because the context was different. Under the Federal Rules of Civil Procedure, expert witness methodology is broadly discoverable. If an expert relied on ChatGPT to form an opinion, opposing counsel has a right to probe how that opinion was built.
The Connecticut ruling didn’t contradict the Texas and New York decisions so much as highlight an important boundary. Work-product protection applies when materials reflect an attorney’s mental impressions during trial preparation. Expert witness methodology occupies different legal territory.
No new AI privilege, just old rules applied to new tools #
Legal observers have noted something important about all three rulings: none of them created a novel AI-specific privilege. Every court applied existing legal doctrines, specifically work-product protection and expert-discovery rules, to a new category of materials.
Future disputes will likely hinge on the same question that has governed work-product doctrine for decades: do the materials reflect an attorney’s mental impressions and litigation strategy? The AI is just a new medium through which those impressions might be expressed.
What remains to watch is whether federal circuit courts will weigh in and whether any appellate decisions will cement or fracture the emerging consensus. The Texas and New York rulings are trial-level decisions, meaning they carry persuasive but not binding authority outside their jurisdictions. The Connecticut ruling came from a federal magistrate, adding another layer of jurisdictional complexity. Until an appellate court addresses the issue directly, the landscape will remain a patchwork, state by state, circuit by circuit, with outcomes turning on whether the AI was a tool of litigation strategy or expert methodology.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our