US: state bar guidance on AI is a patchwork, so how do you comply?
By the ChamberOne team 5 minute read
In the United States, the rules of professional conduct are made state by state. When AI arrived, many state bars and ethics committees began to publish their views, and they did not all say the same thing. A firm with lawyers in several states, or a lawyer admitted in more than one, cannot assume that guidance from one jurisdiction applies everywhere. This article is general information and not legal advice. Opinions are issued, updated and sometimes superseded, so read the current text from each relevant state bar before you rely on any summary, including this one.
Key points
- US conduct rules are state by state, and ethics opinions on AI differ in status and detail.
- Common themes are competence, confidentiality, supervision, verification, candour to the court, client communication and fees.
- A single firm-wide policy built on the strictest reasonable reading is simpler than separate rules per state.
- Track current guidance, court standing orders and client outside counsel terms, and keep an audit record.
Why a patchwork exists
Each state, and often each court, sets its own rules for lawyers. Ethics opinions are issued by bars, committees or courts, with different status in different places. Some are advisory, some carry more weight, and some are guidance rather than rules. Meanwhile, individual judges have issued their own standing orders about AI in filings, which add another layer.
The result is that a lawyer practising across state lines faces a mixed set of expectations, some stated in detail and some in general terms. Differences in wording can matter. The sensible response is to organise around themes, then check each state for its particular points.
Themes that recur
Reading across the published views, certain themes come up again and again. Most are applications of long-standing duties to a new tool, not new duties.
- Competence: a lawyer should understand the benefits and risks of the technology they use, or get help from someone who does.
- Confidentiality: client information should not be put into tools in ways that risk disclosure, and the lawyer should understand how a provider handles data.
- Supervision: lawyers remain responsible for the work of others and of tools, and must supervise accordingly.
- Verification: output must be checked, particularly legal citations and factual statements.
- Candour to the court: filings must be accurate, and false or non-existent authorities are a serious problem.
- Communication: in some circumstances clients should be told about the use of AI, particularly when it affects their matter or the fee.
- Fees: charging for time saved, or for the cost of tools, raises questions about reasonableness.
Where states tend to differ
Differences are often in degree and detail. Look for these points in each state.
- Whether informed client consent is expected for some uses, and for which.
- Whether clients must be told about AI use, and when.
- How fees for AI-assisted work are to be treated.
- What is said about self-learning tools that retain inputs.
- Whether the court has a specific disclosure or certification requirement for filings.
- The weight of the document: formal rule, ethics opinion or informal guidance.
Do not assume that silence means permission. If a state has not addressed a point, general principles still apply, and bars may be watching for cases.
One policy for the strictest reasonable reading
A practical approach for a multi-state firm is to write a single policy that meets the strictest reasonable reading of the themes, then add state-specific notes where needed. That is simpler to train and supervise than a different rule in each office, and it means a lawyer who moves or works across borders does not have to relearn the policy.
The policy should be justifiable in terms of the duties of competence, confidentiality and supervision, rather than tailored to avoid discovery. If a state's guidance demands more, adopt it firm-wide where practical.
- Meets the strictest reasonable reading of the themes
- Simpler to train and supervise
- A lawyer who moves does not have to relearn it
- Harder to train and supervise
- Lawyers working across borders must relearn the rules
What such a policy might contain
- 1
A list of approved tools, with the data terms checked, and a rule that unapproved tools are not used for client matters.
- 2
A rule that client confidential information is never entered into tools without confirmed protections.
- 3
A verification requirement: every citation and quotation checked against an authoritative source by a person, and recorded.
- 4
A supervision rule: partners remain responsible, and juniors disclose AI use to their reviewers.
- 5
A client communication rule: tell clients about significant AI use in the engagement letter or when it affects the matter, and get consent where needed.
- 6
A fees rule: bill for lawyer time actually spent and for value delivered, and explain any technology charges.
- 7
A court rule: check the local rules and standing orders before filing, and comply with any disclosure requirement.
- 8
A record: an audit trail of who used which tool on which matter.
Keeping up with the changes
Because guidance is changing, assign someone to track it. A knowledge lawyer or the general counsel can keep a short table of each state where the firm has lawyers, with links to the current guidance and a note of its status and last check date. Review it every few months, and after any significant development. Subscribing to the bar's communications and to a reliable legal ethics newsletter helps.
- 1Keep a table of each state with its guidance, status and last check date
- 2Review it every few months
- 3Review it again after any significant development
- 4Update the firm-wide policy where guidance demands more
- Then back to step 1.
Standing orders and local court rules
Courts sometimes issue requirements specific to AI, such as certifying that citations have been checked or disclosing that a tool was used. These vary by judge and by court, and they can change quickly. Before filing in any new court, check the local rules and the particular judge's orders. Add this as a line on your filing checklist. A policy that is excellent in general can still fail if a particular judge has a requirement the team did not know about.
Working with clients and their own rules
Corporate clients increasingly have their own AI terms in outside counsel guidelines. Some prohibit AI on their matters, some require notice and some set conditions on tools. Treat these as contractual requirements, in addition to the bar's views, and make sure that the people on the matter know them. A short table of client-specific restrictions, kept where the team can see it, avoids embarrassing mistakes.
What a policy cannot do
A written policy is a starting point. It does not make a firm compliant if staff ignore it, and it does not guarantee that a particular state will agree with the firm's reading. Training, supervision and an audit record turn a policy into practice. When something is unclear, ask the bar's ethics hotline, where one exists, or take advice from ethics counsel. Document what you did and why.
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
Generative AI guidance for lawyers: how six markets compare
How to find what regulators, courts and bar bodies have said in six markets, and the common themes across them.
5 minute readUS: ABA Formal Opinion 512 on generative AI, in plain language
What the American Bar Association's first ethics opinion on generative AI covers, topic by topic.
4 minute readEU AI Act: what it means for firms outside the EU with EU clients
How a firm outside the EU should think about the EU AI Act when it works for European clients.
5 minute read