District of Columbia · state bar
Ethics Opinion 388: Attorneys' Use of Generative Artificial Intelligence in Client Matters
The opinion says generative AI does not alter a lawyer's duties to clients or to the courts, describes these tools as predicting statistically probable text rather than reporting what is in a database, and asks lawyers to understand enough to judge accuracy and confidentiality.
Tracker entry · District of Columbia
- Court. District of Columbia Bar, Legal Ethics CommitteeCOURT
- Level. state barLEVEL
- Date. published April 2024DATED
- What it requires. verification dutyREQUIRES
- Source. Read the documentSOURCE
Last checked · we opened the source on that day and drew the requirement above from its text.
The opinion is unusually clear about the mechanism, which is why it is useful to a buyer. It says these tools are not search engines that report hits on existing data in a constantly updated database. The dataset may be incomplete, out of date or biased, and the system is not built to report what is in it accurately. It is built to create new content by predicting the next word.
From that the committee draws the practical point: statistically probable outputs are not what a lawyer searching for controlling authority needs. Lawyers do not have to be programmers, but they do need to understand enough about how the tool works, what it does, and its limits to be satisfied that the output will be accurate for the task and will not breach client confidentiality. Supervision, candour to the tribunal, fee reasonableness and the client file are named as well.
This entry is descriptive, not legal advice. It records what one document says. Read the order that governs your own division before you rely on any of it, and check whether it has changed since the date above.
Back to every jurisdiction in the tracker, or readhow to check an AI citation before you file.