Technology’s Effect on Advocacy
different case merely because both plaintiffs alleged harassment. Trial lawyers understand this intuitively because valuation is contextual, clients understandably do not. AI can also be highly responsive to the assumptions embedded in a prompt. Consider the difference between: “My employer fired me after I complained about discrimination. How strong is my case?” and “My employer retaliated against me and illegally fired me because I exposed blatant discrimination. How much should they have to pay?” The second question has already decided causation a and liability before the AI begins its analysis. If the system accepts those premises, the resulting answer may reinforce the client’s existing beliefs rather than test them. By the time the client meets with counsel, the lawyer is no longer evaluating the case on a blank slate. We may be negotiating against an invisible co- counsel who has already promised our client a seven- figure recovery, complicating everything from intake to mediation.
Clients may input extraordinarily sensitive information into public AI systems: medical records, psychiatric histories, personnel files, performance reviews, text messages, emails with witnesses, settlement communications (just to name a few). And, potentially, communications with their lawyers. The client may view the interaction as private because it feels private. She is sitting alone with her phone or computer, typing into a box. There is no obvious third person in the room. The confidentiality implications of AI platforms depend on the particular product, account, settings, terms, security measures, and ways in which submitted information is retained or used. California lawyers have already been instructed to take those distinctions seriously in our own use of AI. The State Bar’s current guidance warns attorneys not to input confidential client information into generative AI systems lacking adequate confidentiality and security protections and advises attorneys to understand how a provider uses inputs, including whether information may be shared or used to train or improve a product. Our clients are unlikely to have read the State Bar’s AI guidance and likely will not read the platform’s terms of use, creating an opportunity for basic client counseling. We routinely tell clients not to post about their lawsuit on Instagram. We warn them that text messages may become evidence. We tell them not to delete documents. We prepare them for the fact that defense counsel may examine years of medical history. It may be time to add another instruction: Do not upload documents or confidential case information into public AI platforms without talking to us first.
It is also an important reminder that expectation management needs to begin earlier than it once did.
One useful question during intake m ay now be: “Have you used ChatGPT or another AI tool to research your claims or estimate the value of your case?” The answer can tell us a great deal about what expectations have already been created.
Your Client May Be Feeding the Internet Your Case File
There is another concern that receives less attention from clients because it is considerably less exciting than asking how much their lawsuit is worth: privacy.
40
Consumer Attorneys of California
FORUM September/October 2026
Made with FlippingBook - Online catalogs