Thank you to Smokeball for seeing some of my legal AI related content on socials and asking me to share some insights in their new Community Lounge for AI use. This is a topic that I am passionate about as a lawyer for 20 years who has seen the legal profession transform itself repeatedly during that time. When I started practicing, I used to send faxes. A lot has changed since then, and the biggest transformation is happening around us with the rise of artificial intelligence. AI has produced a variety of reactions in the legal world: excitement, suspicion, fear, curiosity. My own feelings toward legal AI use depends on the day. When I use it to simplify and automate my workflow (having it transcribe audio, or create the gist of an email to a client that I can then revise), I am excited, eager, and curious to explore more features. But when I see lawyers in the news being sanctioned for their AI use, I experience trepidation and sadness.
In Illinois, our ARDC has put out guidance about legal AI use, as have our bar associations. These provide useful frameworks as attorneys grapple with how to incorporate legal AI into their practice and to what extent, amidst an ever-evolving environment. By keeping this framework in mind, lawyers are empowered to evaluate their unique workflows and office processes and determine how to keep practicing ethically and competently with yet another new legal tool for our arsenal.
Accountability is a bedrock of our profession. Let’s all promise each other right now: we will always review anything that AI produces, exactly as if it was a new associate right out of law school. We must review every single case that AI provides – not only that it exists, but that it says what the AI says it says. This is no different than being sent a motion from a brief bank and reading the cases cited in it, because you didn’t write the initial motion yourself from scratch, and a lawyer should always read every single case cited in a motion he files. This is what I was told by an attorney of 40 years when I first started practicing, and it is advice that should always hold true. The AI errors cited during show cause hearings can be caught this way – the hallucinations, the mis-cites, and the misstatement of case law.
We also have a duty of competence. This is where we have to do the legwork. There are so many AI options out there for lawyers, and they all vary. For example, do you know the difference between Narrow AI and Generative AI? Do you know the difference between an open and closed AI environment? Generative AI produces output like text, video, and images. An open AI environment pulls from the entire universe of resources available and trains itself on it; a closed AI environment, like Smokeball’s Archie AI built into each matter, uses only the documents added into the environment, which minimizes hallucinations. Do you use ChatGPT? Which model? Is your legal team using Claude, as a lot of the BigLaw firms are, or are they using GC AI, as a lot of in-house legal teams prefer? What are the models out there, what do they do, how are they trained, who owns them, how do they retain and protect data? Are they secure? Legally compliant? When is their output discoverable?
You can only answer if you know what your AI agent of choice is about. We do not need to have their privacy policies memorized to be competent, but we need to know about the model we are using and what it will do with the information we put into it. This is competence. After all, you would not use any other software tool – or human employee – to help you with your work unless you knew how it was helping you, what its practices were, and what would happen to the work you gave to it. It sounds daunting, but it need not be. We are lawyers, and due diligence is a pillar of our practice.
We also have a duty to protect client confidentiality, and that is even more true with AI. Depending on our area of practice, we possess a great amount of sensitive client information. We have to be aware of what client information we feed into AI, whether AI will use that information for training purposes, how long AI retains the information even after we delete it or restrict access, who else has access to that information through the AI, how the AI firm protects that information from breaches, and more. Many lawyers resolve not to use AI at all in order to avoid this exercise. However, there is no reason to be intimidated; this is the confidentiality and data protection work we already do, but just with a new tool and in a new (digital) dimension. I started my career with knowing where our paralegal kept the small key that unlocked our “secure evidence locker,” or a big filing cabinet where we kept our most sensitive information and discovery. Today, I review my GenAI documentation and make sure that it protects client confidentiality.
This is a very general overview based on how we treat legal AI here in Illinois. Your ARDC and various state bars or committees will likely have produced a tranche of AI resources that will help you learn the environment and how to use AI in a way that fits your practice and keeps you accountable, competent, and protective of sensitive information.
If this kind of content is useful to some of my fellow Smokeball using lawyers and staff, I will post more. I am also hoping to read posts from others in the legal industry about AI because I, too, am always learning.
