The unauthorized-practice line
A tool that applies the law to a person's specific situation and tells them what to do is giving legal advice, which non-lawyers generally may not do. This is the line that separates information, which anyone may share, from advice, which is reserved to those licensed to give it. The distinction is not about the format of the answer but about what the answer does.
This unauthorized-practice line shapes what legal AI may safely offer. A tool can explain what a term means, describe how a process generally works, or point someone to the right form without crossing it. The moment it weighs the particular facts of one person's situation and recommends a course of action, it is doing the thing the rules reserve to lawyers.
The line exists to protect people from unqualified guidance, but it also constrains how far these tools can go. Designers of legal AI have to decide deliberately which side of the line each feature sits on, because the same technology can be used to inform or to advise, and only one of those is generally permitted.
New regulatory experiments
Some jurisdictions are testing new approaches. Regulatory sandboxes, for example in Utah and Arizona, let new legal-service models, including technology-driven ones, operate under supervision so regulators can learn what genuinely expands access versus what exposes people to risk. Instead of banning a new model outright or approving it blindly, the regulator allows it to run within defined limits and watches what happens.
The idea is to gather evidence rather than guess. A sandbox can collect data on whether a tool actually helps people, whether it causes harm, and how often, before any decision is made about whether to allow it more broadly. That turns a hard yes-or-no question into something that can be studied.
These experiments matter because the old rules were written for a world of human practitioners, and they do not map cleanly onto software. A supervised testing ground gives regulators a way to adapt without either freezing innovation or letting untested tools loose on vulnerable people.
Professional duties
When lawyers use AI, professional obligations still apply, competence, supervision of the tool's output, and protecting client confidentiality. Using a tool does not transfer responsibility to the tool. The lawyer remains answerable for the work, just as they would be for the work of anyone they delegated to.
Competence means understanding the tool well enough to use it responsibly, including its limits and its tendency to err. Supervision means checking what it produces rather than passing it on unread. Confidentiality means being careful about what client information is fed into a system and where that information goes.
Good governance keeps a human accountable for what the technology produces. Whether the rule comes from a court, a regulator or a professional body, the through-line is the same: the technology can assist, but a named person carries the duty. That accountability is what keeps the use of AI tethered to the protections the legal system is built on.
What this is and is not
This is research and educational material on the regulation and governance of legal AI. It is general legal information, not legal advice, and it does not describe any individual's matter.