AI in Criminal Defense: A Field Guide for Practicing Lawyers

This is a continuing legal education program for Minnesota attorneys on using artificial intelligence competently and ethically in an active practice. It is built around five questions every lawyer should be able to answer before bringing these tools into their work — what AI actually is, what competence now requires, how to protect client confidentiality, how to verify AI output, and where to start. It is written for working lawyers, not technologists, and it is grounded in the rules we already practice under.

The rules of professional conduct already speak to this. Comment 8 to Rule 1.1 has, since 2015, defined competence to include keeping abreast of "the benefits and risks associated with relevant technology." In July 2024, ABA Formal Opinion 512 — the first formal ethics opinion on generative AI — concluded that generative AI is exactly that kind of relevant technology. The obligation is not to become a computer scientist. It is, in the words of Opinion 512, to have "a reasonable understanding of the capabilities and limitations of the specific tools" you use. This program is about reaching that reasonable understanding.

Who This Is For

Solo and small-firm practitioners, public defenders, and any attorney who suspects the profession is changing faster than the profession is admitting. It is deliberately practical: less about how large language models work under the hood, more about what it looks like to use one well on a real file, on a deadline, without violating a single rule of professional conduct.

It is also honest about the risks. Every cautionary tale you have heard — the fabricated citations, the sanctions, the lawyer who filed a brief full of cases that did not exist — is real, and this program treats those failures as what they are: failures of supervision, not indictments of the technology. A tool that occasionally invents a citation is a tool that must be verified, the same way you would verify the work of any junior associate. The rule is not abstinence. The rule is supervision.

The Five Questions

1. What is AI actually — and what isn't it?

Stripped of jargon, a large language model is software that has read an enormous amount of human writing and learned to produce language that continues from any prompt. It is not searching a database and it is not conscious. It is a very well-read pattern engine. Understanding that plainly — what it does, what it does not, and where it fails — is the foundation for everything else, and it is most of what "reasonable understanding" under Opinion 512 actually requires.

2. What does competence require — in both directions?

Competence now runs two ways. There is the competence to use these tools when they serve the client, and the competence to supervise them so they never harm the client. Rule 1.1 and Opinion 512 do not require you to adopt AI. They require you to understand it well enough to know when it is appropriate, what its limits are, and how to check its work. The hallucination problem is real — and it is entirely manageable by a lawyer who verifies before filing.

3. What's the plan for client confidentiality?

This is where most of the genuine risk lives, and it is manageable with a written plan. The one path to avoid is pasting client information into a free public chatbot whose terms may permit training on your input. The paths forward: disclose AI use and obtain informed client consent in your engagement letter; use enterprise accounts whose contracts explicitly prohibit training on your data; or, for the most sensitive work, run models locally on your own hardware so the data never leaves your machine. Pick one. Write it down. Be able to answer "how do you handle client data when you use AI" with specifics rather than hope.

4. How do I verify what AI produces — every time?

Every case citation, every factual claim, every legal conclusion gets checked against a primary source before it is used — the same standard you would apply to a memo from a first-year associate. This is not a burden the technology creates; it is the supervision the rules already require. The lawyers who have been sanctioned did not get there because AI failed. They got there because they filed work they had not read.

5. What's the right first step for my firm?

Deliberately small. Pick one tool — not three. Run it on one repetitive task you already do every day, the boring one you would pay to skip. Do it that way for thirty days, then judge. Most lawyers try it twice, get one bad answer, and quit. The on-ramp is a single tool, a single task, and enough patience to get past the first week.

The Live Demonstration

The program includes a live walkthrough on a fictional case file — a synthetic Minnesota second-degree murder matter built entirely for teaching, with no real persons or facts. The point is not to show that AI is impressive. It is to show the division of labor: the lawyer brings the judgment, the context, and the right questions; the tool brings synthesis, speed, and tireless first-draft work. The lawyer asks. The AI executes. The lawyer verifies. That is the model, and seeing it run on a real-looking file makes it concrete in a way slides cannot.

Bringing This Program to Your Organization

This CLE is available to bar associations, law libraries, public defender offices, and firms. It is designed to send attorneys home able to answer all five questions and take a concrete first step — not merely to have watched a presentation about a trend.

If you would like to discuss bringing this program to your group, or want to talk through applying any of it to your own practice, grab 20 minutes on my calendar or email travismkeil@keildefense.com.

Travis Keil, Keil Defense — Chanhassen, Minnesota. To bring this CLE to your organization or to talk through your own practice, book a time, email travismkeil@keildefense.com, or call 651-315-3097.

Note: This page describes a continuing legal education program and is intended for attorneys. It is not legal advice and does not create an attorney-client relationship. References to the Minnesota Rules of Professional Conduct and ABA Formal Opinion 512 are provided for professional education; lawyers should consult the current text of the rules and opinions directly.