Texas and New York courts shield AI prompts from discovery
Two state courts ruled in June 2026 that AI prompts and outputs created during litigation preparation are protected work product, shielding them from discovery demands and establishing a concrete legal precedent for AI-assisted legal work.
The Texas Business Court, 11th Division, held on June 3, 2026, in *Tate Group Automotive, LLC v. Legacy Automotive Capital, LLC* (No. 25-BC11B-0020) that a litigant's ChatGPT conversations qualified as protected work product under Texas Rule of Civil Procedure 192.5 Reuters. One day later, on June 4, 2026, the Supreme Court of New York, Nassau County, issued a decision in *Assini v. Hayward* (2026 NY Slip Op 26086) that quashed non-party subpoenas seeking AI prompts, uploads, and outputs, recognizing AI-assisted litigation preparation as protected trial-preparation material under New York's work-product doctrine Reuters.
Work-Product Doctrine Extends to Generative AI
Both rulings apply established attorney work-product protections—legal doctrines that shield materials prepared in anticipation of litigation from mandatory disclosure—to generative AI outputs. The Texas court applied the state's discovery rules; the New York court applied the state's trial-preparation protections. Neither ruling imposed criminal penalties or regulatory fines; both were discovery disputes in civil litigation Reuters.
The decisions clarify that AI prompts and outputs created for litigation are trial-preparation materials when generated by counsel or their team in anticipation of litigation, and therefore qualify for the same confidentiality protections afforded to attorney notes, strategy documents, and other work-product materials.
Implications for Agent-Driven Legal Practice
These rulings carry direct implications for the emerging agent economy in legal services. As law firms deploy AI agents to conduct research, draft motions, analyze discovery, and prepare case strategy, these decisions establish that agent-generated materials—prompts fed to the agent, outputs it produces—can retain privilege if created in the course of litigation preparation. This protection extends the historical boundaries of attorney-client privilege and work-product doctrine into the domain of autonomous AI systems used by legal teams.
The rulings do not, however, extend blanket protection to all AI use in law firms. Courts distinguished between AI outputs created for litigation versus those created for business operations or general counsel work. Only materials generated in anticipation of dispute or trial qualify for protection under these decisions.
Both cases were reported by Reuters in August 2026, months after the rulings, as the precedent began circulating among legal practitioners and legal technology vendors building agent-assisted practice management tools.