Announcing Free Law Project's AI Policies
A volunteer uses an AI assistant to write code for CourtListener. A staff engineer asks a chatbot to untangle a stubborn bug. A researcher builds a system that reads thousands of court opinions. Artificial intelligence is already woven into how legal work gets done, including ours.
We think the people who rely on our work deserve to know how we handle that. So today we're publishing the three policies that guide how Free Law Project develops and uses AI: in the open, for anyone to read and build on.
Why we wrote these down
We work with the law, and in the law, mistakes have consequences. A made-up citation or a garbled opinion doesn't just embarrass us: it can mislead a self-represented litigant, a journalist on deadline, or a legal aid attorney with a full caseload. Those are the people our work exists to serve.
At the same time, AI genuinely helps a small nonprofit team do more, and we build AI-powered tools ourselves. So rather than pretending AI isn't here, or adopting it without guardrails, we wrote down how we want to use it. The guiding principle running through all three policies is simple: a person must stand behind every work product. AI is a tool, not a contributor.
Three policies, three audiences
-
AI Policy for Use of Third-Party AI Tools and Systems: This policy is for staff who use tools like Claude and ChatGPT in their daily work. It requires staff to verify AI outputs before relying on them, protects sensitive information, and rules out tools that train on our data without an opt-out.
-
AI Policies for Development of AI Systems: For developers who build AI systems at FLP. It puts accuracy ahead of speed, requires us to document design choices, insists that AI-generated content be labeled as such, and scales safeguards to the stakes involved.
-
AI Policies for Volunteers Working with Free Law Project: For volunteers who contribute code, data, and research. It asks volunteers to understand and disclose their AI use, handle data carefully, and tell us when AI tools introduce errors, so we can all learn from them.
Why they're public
Transparency is part of our DNA: our code is open, our data is open, and our policies should be too. Publishing them keeps us accountable to the community we serve.
But we also hope these policies spark something beyond Free Law Project. Every organization is facing the same questions about AI right now, and few have the time to answer them from scratch. If our policies give you a head start on writing your own (as a starting point, a template, or just proof that it can be done), that's a win for everyone. Take them, adapt them, make them yours.
Read, subscribe, and make them better
AI is moving fast, and these policies will evolve with it. We encourage you to read the policies and subscribe to them on our wiki so you're notified when they change. And if something seems off, unclear, or missing, tell us. These policies will be better for it.
Thank you
These policies didn't come together overnight. They're the product of months of research, drafting, and revision across many teams, and we're grateful to everyone who shaped them.
Our deep thanks go to the Samuelson Law, Technology & Public Policy Clinic at Berkeley Law: Erik Stallman and his students Ainsley Rosenthal, Eric Miranda, and Sara Qi, whose research and drafting shaped these policies from the very first conversation. Thanks to our own staff, whose candid feedback made every policy sharper and more practical. And thanks to Andrew Nordick with Mitchell Hamline School of Law, who first prompted the discussion that we needed AI policies at all.
Building legal technology carefully, and in the open, is what we do. If that mission matters to you, support our work, or start by reading the policies and telling us what you think.
