Cases on AI tools and attorney-client privilege / legal professional privilege
The UK’s Upper Tribunal held in Munir v. Secretary of State for the Home Department that an immigration lawyer who used an AI system to draft legal submissions — without disclosing this or verifying the output — had waived legal professional privilege over the AI-generated content. The ruling is England and Wales’ first direct holding on AI tool use and privilege waiver.
A federal court in Jones v. Delta prohibited counsel from using AI tools to prepare deposition questions or analyse deposition transcripts in real time, finding that AI-assisted deposition preparation created unfair informational asymmetry. The ruling is one of the first to restrict AI use specifically during the deposition phase of litigation, rather than in written pleadings or research.
In Tym v. Cerno, a California federal court confirmed that the specific prompts a lawyer types into an AI system can themselves constitute protected opinion work product, because they reveal counsel’s mental impressions and litigation strategy. The ruling extends classic work-product doctrine into the AI era, shielding lawyer-AI interactions from compelled disclosure in discovery.
A Kansas federal court in Jeffries v. Harcros barred defence counsel from using AI tools trained on or containing any materials produced during discovery — an unprecedented restriction protecting the confidentiality of litigation-sensitive documents. The ruling signals that courts will impose technology-specific use restrictions where AI systems risk absorbing and reproducing protected discovery materials.
In Morgan v. V2X, a Virginia federal magistrate ordered detailed logging and segregation protocols for AI tools used during litigation, treating AI-assisted legal work like any other potentially privileged electronic data. The ruling establishes that parties using AI in active litigation must be prepared to identify, log, and defend every AI-generated work product claimed as privileged or protected.
A Texas federal court held that ChatGPT-generated legal research used by defendant’s counsel during discovery was protected as opinion work product — provided it was not shared with opposing parties. Warner v. Gilbarco is the first US ruling to squarely hold that a lawyer’s AI-assisted analytical work can qualify for work-product protection under Federal Rule of Civil Procedure 26.