Lesson objective
Explain the professional duties that make AI training necessary, and set a session goal of competent practice rather than awareness of risk.
Plain-language explanation
Why this matters now
An instructor needs a clear answer to a question participants rarely ask aloud: why am I here? The answer shapes everything else. If training exists to make lawyers aware of AI risk, a lecture with a list of risks does the job. If training exists because lawyers have a professional duty to be competent in what they use, and because firms have duties to supervise, then awareness is not the goal. Competent practice is.
The duties are real, and they must be taught with their caveats intact.
Law / rule: Comment 8 to ABA Model Rule 1.1 (6 August 2012) says that to maintain competence a lawyer should keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology. Note the words "should" and "relevant technology". Most, but not all, states have adopted a version of the comment; check the jurisdiction of your audience before saying it applies to them.
Law / rule: ABA Model Rules 5.1 and 5.3 address the responsibilities of managerial and supervisory lawyers for other lawyers and for non-lawyer assistance. They are model rules only. Each jurisdiction adopts its own, with different numbering and wording.
Professional guidance: ABA Formal Opinion 512 (29 July 2024) applies competence, confidentiality, communication, candour, supervision and fees to generative AI tools, including a lawyer's understanding of a tool's capabilities and limitations, and the role of firm policies. It is guidance interpreting the Model Rules and does not bind a lawyer by itself.
A question to sit with
Imagine a participant who can list every risk from your session perfectly a week later, and who still pastes a client's draft agreement into an unapproved tool on Friday afternoon. Did your session succeed?
What goes wrong
1. Awareness is mistaken for the goal. The session is judged a success because participants can name the risks. Naming a risk and managing it in a live document are different abilities.
2. The duty is taught as a threat. "You could be disciplined." Fear may get attention; it does not show anyone what to do instead. Module 7 treats this in depth.
3. The duty is overstated. The instructor says "every lawyer is required to be competent in AI". The model comment says "should", speaks of "relevant" technology, and has not been adopted everywhere. A lawyer who knows this stops trusting the instructor.
4. Only prohibitions are taught. The session lists what not to do and never shows the approved route. Participants who still have the work to do find their own route.
5. Supervision is left out. The session addresses the individual lawyer's use and ignores that partners supervise associates, paralegals and assistants whose AI use they may never see.
The practical answer
AI Coalition Network recommendation: write the session goal as a competent act. A competence goal names what a participant will be able to do, to what standard, in which situation. An awareness goal names what they will know. This is an educational recommendation, not a legal requirement.
| Awareness goal | Competence goal |
|---|---|
| Understand the confidentiality risks of AI tools | Decide, for a given piece of client material, whether the firm's policy permits it to go into a given tool, and route it correctly |
| Be aware that AI can hallucinate citations | Verify every authority in an AI-assisted draft against the source before it leaves the building, and record that it was done |
| Know the ethics guidance on AI | Identify which duties a proposed AI use engages, name the jurisdiction whose rules apply, and find the current adopted rule |
| Recognise supervision obligations | Review a junior colleague's AI-assisted work as a checker, not only as an approver, and record the review |
Best practice: teach the duty as the reason, and the routine as the content. Spend two minutes on where the duty comes from, with every caveat. Spend the rest of the time on the routine that discharges it. A lawyer who understands why a check exists and has performed it on a realistic document has at least rehearsed the moment they will meet on Monday. A lawyer who has only heard the rule read out in full has not.
Best practice: teach the approved route alongside every prohibition. For every "do not", the session must answer "then what?".
- Do not put client material into an unapproved tool. Then: here is the approved tool, here is what it may be used for, and here is who to ask when you are unsure.
- Do not rely on an AI-generated citation. Then: here is the research service, here is the check, and here is where you record it.
- Do not bill for time you did not spend. Then: here is how the firm records AI-assisted work, and here is who to ask about your jurisdiction's guidance on fees.
Best practice: include the supervisors. A session for associates should have a companion for the partners who review their work. Supervising lawyers need a different routine: how to ask whether and how AI was used, what to check in work they did not draft, and how to read as a checker rather than as an approver. Teaching only the people who draft leaves the review stage untrained.
Best practice: be precise about "relevant technology". The model comment does not require every lawyer to use AI, and nothing in this programme should suggest it does. It speaks of keeping abreast of the benefits and risks of relevant technology. A lawyer who chooses not to use a tool still needs to understand enough to supervise colleagues who do, to answer a client who asks, and to recognise AI-generated material from an opponent. That framing is accurate and it reaches the sceptic who has no intention of adopting anything.
Remember this
- Training exists to build competent practice, not awareness; a participant who can list risks and still makes the Friday mistake has not been served.
- Teach Comment 8 to Model Rule 1.1 with "should", "relevant technology" and the fact that not every state has adopted it; teach Model Rules 5.1 and 5.3 as model provisions.
- ABA Formal Opinion 512 is guidance interpreting the Model Rules, not binding law.
- Every prohibition needs an approved route alongside it.
- Write session goals as competent acts, and train the supervisors as well as the people who draft.
Legal example
The fictional firm Garrowby Leith LLP asks its practice support lead, Ines Montcalm, to update its annual AI session for trainee lawyers and paralegals. The previous year's goal read: "Participants will understand the ethical risks of generative AI." The session consisted of a slide quoting the model technology competence comment with the words "every lawyer must", a list of eleven prohibited practices, and a warning about discipline. Ines rewrites the goal as three competent acts: decide whether a given document may go into the firm's approved tool under the firm's policy; verify every authority in an AI-assisted draft before it goes out and record the check; and prepare a short note to a supervising lawyer describing how AI was used on a task. She replaces the competence slide with the text of the comment, highlighting "should" and "relevant technology", and a note that most but not all states have adopted a version, followed by the adopted rule in the firm's own jurisdiction, which she confirms with the firm's general counsel. She reduces eleven prohibitions to four, and pairs each with the approved route and the name of the person to ask. She runs a separate forty-minute session for supervising partners on reviewing AI-assisted work as a checker. When she mentions the Thomson Reuters Institute figure on unsanctioned tools, she gives its date, its sample and the fact that it is industry research, and she uses it only to explain why every prohibition in the session now has a route beside it. Three months later the firm's general counsel notes that questions about which tool may be used for what have moved from the IT helpdesk to the named contacts in the session.
Responsible-use warning
Never overstate a professional duty to motivate a room. Comment 8 to Model Rule 1.1 uses "should", speaks of "relevant technology", and has not been adopted in every state; Model Rules 5.1 and 5.3 are model provisions adopted differently by each jurisdiction; ABA Formal Opinion 512 is guidance, not law. Never present AI Coalition Network's named-reviewer standard as a legal requirement. And never teach a prohibition without the approved route beside it.
Human-review requirement
A licensed attorney reviews the session's statement of duties before delivery and owns its accuracy. The reviewer confirms that each rule is identified as a model provision or as the adopted rule of a named jurisdiction, that the adopted rule for the audience's jurisdiction has been checked against its current text, that guidance is dated and attributed, that firm policy is distinguished from legal duty, and that every approved route named in the session matches the firm's current policy and tools.
Practical exercise
Turn an awareness session into a competence session. Use a fictional firm and fictional participants.
- Write down the goal of an AI session you have delivered or attended, or use: "Participants will be aware of the risks of AI in legal practice."
- Rewrite it as three competent acts, each naming what the participant does, to what standard, and in which situation.
- Write the two-minute statement of the duties behind the session. Include Comment 8 to Model Rule 1.1 with its "should" and "relevant technology" wording and the adoption caveat, and describe Model Rules 5.1 and 5.3 as model provisions.
- Name the jurisdiction of your fictional audience and write where you would check its adopted rule. Do not state the content of any rule you have not checked.
- List every prohibition in the session. For each, write the approved route and the person to ask.
- Design a companion session outline of no more than five lines for supervising lawyers.
- State the AI Coalition Network named-reviewer standard in one sentence that makes clear it is an educational standard, not a legal requirement.
- For one competence goal, describe in two sentences what you would see a participant do in the session that tells you they can perform it.
Knowledge check
-
What is the difference between an awareness goal and a competence goal?
Answer guidance: An awareness goal names what a participant will know; a competence goal names what they will be able to do, to what standard, in which situation, such as verifying every authority in an AI-assisted draft before it leaves the building and recording the check.
-
Which words in Comment 8 to Model Rule 1.1 does an accurate instructor draw attention to, and why?
Answer guidance: "Should" and "relevant technology", because the comment does not require every lawyer to use AI; together with the fact that not every state has adopted it, they stop the instructor from overstating the duty.
-
What can an instructor responsibly take from the Thomson Reuters Institute figure that 34% of respondents used unsanctioned tools?
Answer guidance: A design question rather than a conclusion: a session that teaches only prohibitions leaves people without a route when work still has to be done. The figure is industry research from a 2026 multi-country survey and does not explain why anyone used such tools.
-
Why should AI training include supervising lawyers as well as those who draft?
Answer guidance: Because supervisors review work they did not draft and may not know how AI was used; they need their own routine for asking, checking and reading as a checker, and the model supervision rules address their responsibilities.
Sources and further reading
- American Bar Association — Comment on Model Rule 1.1: Competence (Comment 8, benefits and risks of relevant technology) (2012-08-06). Accessed 2026-09-14.
- American Bar Association — Model Rules of Professional Conduct. Accessed 2026-09-14.
- American Bar Association Standing Committee on Ethics and Professional Responsibility — Formal Opinion 512: Generative Artificial Intelligence Tools (2024-07-29). Accessed 2026-09-14.
- Thomson Reuters Institute — Future of Professionals Report 2026 (Legal report) (2026-06-22). Accessed 2026-09-14.
- AI Coalition Network — Certification Educational Standard: Human Review of AI-Assisted Professional Work. Accessed 2026-09-14.