The concern most family lawyers raise about AI is whether it is safe to use in a legal practice.
That concern is legitimate, but the reasoning behind it is usually incomplete.
Safety in this context is not a question about the technology itself. It is a question about how the technology is set up, what account you are using, and what governance you have around it.
Handled correctly, AI is safe to use. Handled without any of that infrastructure, it is a real problem.
We have spent a lot of time at Sterling working through this specifically because we want the productivity gains, and we do not want the ethics exposure. Our team walked me through the key pieces.
The Account Type Is the Whole Ballgame
There are two categories of accounts across every major AI provider — consumer-grade and enterprise-grade. This distinction matters more than any other single decision you will make about how AI operates inside your firm.
Consumer accounts are what most people sign up for by default.
They are free or low-cost, easy to set up, and come with terms of service that generally allow the provider to use your inputs to train their models.
If a member of your team is putting client information into a consumer account, that information can become part of the public training data. That is an ethics violation that puts your license at real risk.
Enterprise accounts are structured under a different contract entirely. The data stays yours. The provider is contractually obligated not to use it for training.
You get additional controls, additional tooling, and a very different risk profile.
The critical piece our team emphasized is that it is not enough for the firm to have an enterprise account.
Every person on your team has to be using that account and only that account.
If an attorney or paralegal is running client questions through the free version on their phone or personal computer, the firm still has exposure.
The account has to be the only account they use for firm work.
Governance Is the Piece Most Firms Skip
Getting the accounts right is step one.
Governance is what makes it hold up over time.
Our team’s framing was that a law firm owner cannot supervise every prompt every team member runs, but the absence of any oversight framework is negligent in its own way.
A functional governance setup has three components:
- A written AI policy specifying which accounts and tools are approved and how AI is allowed to be used.
- A signed acknowledgment from every attorney and staff member that they understand and will follow it.
- An ongoing audit process where firm leadership periodically reviews how AI is actually being used across the team.
Sterling’s approach is that every team member’s approved account is accessible to firm leadership for review. Working groups meet regularly to discuss new use cases and to keep the policy current as the technology evolves.
This is not a document you write once and put in a drawer. The technology is changing too fast for that.
Client Consent Belongs in the Agreement
One piece worth calling out separately is how AI use gets addressed in your client agreements. If a client opts out of having AI used on their file, that has to be documented clearly and honored throughout the case. That means your policy needs a mechanism for handling opt-outs, and your team needs to know how to comply when one is requested.
Where to Start if You Have Guidance and Where to Start if You Do Not
The ABA issued Formal Opinion 512 in July 2024 as the first national guidance on generative AI in legal practice. It has not been updated because it was written at the architectural level rather than as a set of specific rules.
If your state has issued its own opinion, use that as your primary reference and treat the ABA opinion as a secondary layer.
Illinois, Texas, Florida, New York, and California all have opinions worth reading.
If your state has not issued anything specific, the ABA opinion is your starting point.
What This Enables
The reason to work through all of this is that once the infrastructure is in place, real productivity gains open up that were not accessible before.
Our team has been using AI extensively for trial preparation — generating question sets, evaluating them against the full case file, and iterating quickly through different angles. The attorney still directs everything and audits every output. The tool speeds the process up significantly and gets the attorney to the strategic work faster than working manually would allow.
That is the pattern worth internalizing. AI does not replace attorney judgment. It handles the repetitive front end of the work so the attorney can spend more time on the parts of the case that actually require legal skill. Get the account structure right, get the policy right, get the ongoing governance right, and the productivity gains follow. Skip any of those steps, and you are one bad prompt away from a very difficult conversation with your state bar.
