OpenAI’s ChatGPT, the human-like chatbot that can write essays, book restaurant reservations and explain complex topics, has been a bit of a one-way mirror—incredibly accessible to users and yet fairly opaque when it comes to explaining how it exactly works.

That has raised some concerns in the legal industry, especially as several legal tech vendors have now announced they were leveraging OpenAI’s GPT-4 model in their offerings and firms are still debating whether they need to issue policies around its use.