US courts begin generating case law on AI-generated legal research and attorney-client privilege as lawyers warn clients that AI chat logs could be disclosed in litigation
A growing body of US case law is prompting urgent client advisories from major law firms about the risks of using AI tools in legal research and privileged communications, with courts beginning to rule on whether AI-assisted work product and chat logs retain attorney-client privilege (the legal rule protecting confidential communications between a lawyer and their client from compelled disclosure). Firms including O'Melveny & Myers and Debevoise & Plimpton have published guidance warning that AI chat transcripts — particularly those generated on third-party 'open' AI platforms — may not be protected from disclosure in litigation because privilege attaches to communications with lawyers, not to interactions with AI tools. The emerging consensus from firm advisories is that closed AI systems (private, enterprise-grade tools deployed within a firm's own environment) offer stronger privilege protections than consumer-facing platforms, though even those protections remain largely untested in court. Debevoise has advised clients that AI research is more likely to attract privilege protection if conducted at the direction of a lawyer, recommending that users include a statement in their AI prompts confirming the research is being done under legal direction. Separately, AI use clauses are increasingly appearing in law firm–client engagement contracts, with at least one firm — Sher Tremonte — including express disclosure provisions about privileged communications and AI use in its retainer agreements. The development has direct implications for how UK and US law firms manage AI risk, given that London offices of US firms are subject to both Solicitors Regulation Authority (SRA) and US bar obligations in respect of competence and confidentiality.
Why this matters
The emergence of AI privilege case law creates an immediate compliance risk for any law firm whose fee earners use AI tools without a defined governance framework — and generates direct demand for legal advice on AI governance policies, vendor due diligence, and client contract terms. For UK-facing firms, the SRA's competence framework requires solicitors to understand the tools they deploy, meaning ad hoc AI use without oversight protocols now carries professional regulatory risk in addition to litigation exposure. The 'why now' is the maturation of AI tool adoption across the profession: firms that were early adopters in 2024–25 are now encountering the downstream legal consequences in live proceedings. The practice area most immediately activated is technology and data, followed by litigation (disclosure obligations) and professional regulation — a combination that uniquely touches most City practice groups simultaneously.
On the Ground
A trainee in the AI and technology practice would assist with drafting AI governance policy documents for law firm clients seeking to formalise their use of generative AI tools, and would prepare vendor due diligence questionnaires assessing whether a proposed AI platform meets the firm's data security and privilege-preservation requirements.
Interview prep
Soundbite
AI privilege case law is making 'open' AI tools a litigation liability — firms that haven't audited their AI stack now face disclosure risk.
Question you might get
“How would you advise a client who has been using a consumer-facing AI tool to draft privileged legal strategy documents, and what steps should their legal team take if those AI chat logs are now subject to a disclosure request in litigation?”
Full answer
Courts are beginning to rule on whether AI-generated research and chat logs can be compelled in discovery, prompting major firms including O'Melveny and Debevoise to issue urgent client guidance on privilege risks. The core issue is that attorney-client privilege protects lawyer-to-client communications — not interactions with a third-party AI system — meaning that chat logs from consumer AI platforms may be entirely unprotected. This matters for law firms because it triggers obligations under both the SRA competence framework in the UK and US bar rules on confidentiality. The broader trend is that AI adoption has run ahead of the legal framework governing it, and courts are now filling the gap. I'd expect UK courts and the SRA to issue analogous guidance within the next 12 months as the volume of AI-assisted work product in English litigation increases.
My notes
saved