Most firms are still deciding their position. A growing number are finding their clients decided first.
Start a conversation with the AI Adoption Concierge, already scoped to client disclosure & engagement letters. Pick a starting point, or describe your situation directly.
Whether AI use must be disclosed is genuinely unsettled as an ethics question and increasingly settled as a commercial one. ABA Formal Opinion 512 addresses communication among the duties generative AI engages, and recommends disclosure in engagement letters in some circumstances — notably where AI use bears on the basis of the fee, or where confidential client information will be processed by the tool. Separately and often sooner, institutional clients have begun writing AI provisions into outside counsel guidelines, covering disclosure, permitted uses, confidentiality and sometimes pricing. Firms frequently discover the question has been answered for them, in a document they signed.
Different clients care about different ones, and few care about all six.
The baseline statement, increasingly ordinary rather than remarkable.
Research, drafting, review, summarisation — clients distinguish between these more than firms expect.
Whether their material goes into a tool, and on what terms. The concern most institutional clients lead with.
The verification process, which reassures far more than any statement about the tool.
The trigger ABA guidance identifies most directly.
A route for the client to restrict use, which is better offered than discovered.
How firms handle it.
Mostly the relationship. The ethics exposure is real but the commercial exposure arrives sooner.
Many firms are debating a disclosure policy while already bound by client documents that address AI use, permitted tools and confidentiality. The audit usually takes a week and frequently settles the debate.
The guidance points at circumstances rather than a universal rule, and the two it identifies most clearly are where AI use bears on the basis of the fee and where confidential client information is processed by the tool. Beyond that, many firms have concluded that a standard engagement-letter clause is simpler than deciding client by client — it is disclosed once, it is consistent, and nobody has to make a judgement call under time pressure. That is a policy choice rather than a requirement, and it should be made with the firm's own counsel.
Far less than firms fear, and specifically. Institutional clients are rarely opposed to AI in principle — many use it themselves and some expect the efficiency to reach their bills. What they object to is their confidential material going into tools that may retain or train on it, and to paying hourly rates for work that no longer takes those hours. Both are answerable. A firm that leads with its data-handling position and its verification process usually finds the conversation short.
Establish what they actually mean before agreeing, because blanket language is often broader than the intent — few clients mean to prohibit spell-check, e-discovery analytics or a legal research platform, and many such tools now have AI components. Then determine whether the firm can comply operationally, which requires knowing where AI sits in its own stack. Agreeing to a restriction the firm cannot honour is worse than negotiating a narrower one, and the negotiation is usually available.
The engagement letter is where ABA guidance points and where it is most durable, because it is signed, dated and retained. Some firms supplement it with a published AI statement, which is useful for RFPs and panel processes where clients increasingly ask. What works poorly is disclosure that first appears on a bill or in response to a question — by then the client is asking because something prompted them to, and the conversation starts from a worse place.
Describe your client base and what your engagement letters say now. The Institute will help you work out a position.