Using AI in a law firm under ABA Formal Opinion 512
ABA Formal Opinion 512, issued July 29, 2024, says lawyers may use generative AI if they understand the tool, protect client information, tell clients when it matters, check the output, supervise its use and bill fairly. A private Mike setup is built to make each of those duties easier to meet, not to replace the lawyer's judgment.
Updated
This page summarizes a published ethics opinion so firms can plan a setup. It is not legal advice. State bars issue their own guidance, and your state's rules govern your practice.
Competence: Rule 1.1
The opinion says a lawyer does not need to be an expert in AI but must have a reasonable understanding of what the tool can and cannot do, and must check its output to a degree that fits the task. Relying on an answer without review is not consistent with the duty of competence.
Every answer Mike gives links to the page of the document or the case it relied on, so checking is a click rather than a search. Each setup includes a working session on the firm's real matters, where attorneys see where the tool is strong and where it needs closer review.
Confidentiality: Rule 1.6
The opinion asks lawyers to understand who can see information entered into an AI tool and whether the tool learns from it. It notes that putting client information into a tool that trains on it can require the client's informed consent, and that boilerplate in an engagement letter is not enough.
In a Mike setup, documents stay in the firm's own installation, and the firm chooses the model provider under its own account, or runs a model on its own hardware. Mike does not train a model on firm documents. The firm still owes its own review of the provider's terms; our security page explains each piece.
Communication: Rule 1.4
Some uses of AI must be disclosed to the client: when the client asks, when the engagement terms require it, or when the output feeds a significant decision in the matter. The opinion leaves much of this to the lawyer's judgment about what the client would reasonably want to know.
Because Mike shows which documents and cases each answer relied on, a lawyer can explain how work product was prepared in concrete terms.
Fees: Rule 1.5
A lawyer billing hourly may charge only for time actually spent, including time spent reviewing AI output, not for the hours the tool saved. General overhead, including learning a tool the lawyer will use across matters, is generally not billable to a client. Charging a client for a tool's cost directly needs a reasonable basis and, where required, the client's agreement.
Mike does not change a firm's billing practices. The firm sets its own policy, and the opinion is a sound starting point for it.
Candor and meritorious claims: Rules 3.1, 3.3 and 8.4(c)
Filing a brief with invented citations or misstated holdings can violate the duties of candor to the tribunal and the rule against frivolous claims. The opinion treats careful review of AI-drafted filings as part of those duties.
Mike checks case-law citations against CourtListener and reads the opinion before it summarizes a holding. That narrows the risk; it does not remove the lawyer's duty to read the authority before relying on it.
Supervision: Rules 5.1 and 5.3
Managing lawyers should set clear firm policy on the use of AI and train lawyers and staff on it, and supervising lawyers remain responsible for the work of the people and vendors they oversee.
In Mike, the firm's administrator decides who has an account, and work is shared by project, so supervisors can see what was asked and what came back. Cuanto Labs is a vendor under that supervision, and the setup agreement states what we can access.
Common questions
Is it ethical for lawyers to use generative AI?
ABA Formal Opinion 512 says yes, if the lawyer meets existing duties while doing it: understand the tool, protect client information, communicate with the client where it matters, check the output, supervise its use and bill fairly. The opinion applies the Model Rules rather than creating new ones, and state bars may add their own guidance.
Do we need client consent to use AI on a matter?
It depends on how the tool handles client information. The opinion says informed consent may be required before putting client information into a tool that could disclose it or learn from it, and that generic consent language in an engagement letter does not satisfy that. A private installation with a provider your firm has vetted changes the analysis, but the decision remains the firm's.
Can we bill clients for time saved by AI?
No, not on an hourly basis. Under the opinion, a lawyer billing by the hour charges for time actually spent, including time spent reviewing the AI's output. Flat or contingent fees are judged by the usual reasonableness standard. Time learning a general tool is overhead and generally not billed to clients.
How do we avoid filing hallucinated cases?
Read every authority before relying on it. Mike helps by checking case-law citations against CourtListener and by linking each statement to the opinion it came from, so review is fast. Treat any unverified citation as unconfirmed, and keep a firm rule that nothing is filed without an attorney reading the cited cases.
Does our state have its own AI guidance?
Many state bars have issued opinions or practical guidance on generative AI, and they differ in detail. Check your state bar's ethics opinions alongside ABA Formal Opinion 512, and follow the stricter rule where they differ.
Sources
Related
Bring Mike to your firm.
Book a 30-minute call. Tell us about your practice and the systems your documents live in, and we'll come back with a setup plan for your firm.
Loading the calendar…
Open the booking page