ABA Formal Opinion 512
July 2024
Lawyers’ use of generative AI: competence, confidentiality, communication, supervision and fees.
In OrviDeskAttorney approval on every matter, disclosure to prospects, and a record of each step.
Security & ethics
The safeguards aren’t a policy document. They’re the order the software runs in, and it can’t run any other way.
Names are checked before a prospect tells the story, because what a prospective client shares can create duties for the firm (ABA Model Rule 1.18).
Lexi says she’s an AI and can’t give legal advice before she asks a single question.
No matter opens, and nothing beyond an acknowledgement goes out, until an attorney decides.
Ruth checks every statement against what the prospect actually said and strikes conclusions nobody made.
Every firm’s records are separated inside the database itself, so one firm’s session can’t read another firm’s data.
Each step records what came in, what was done, and who it went to next.
Ethics guidance
Guidance changes and differs by state. Your firm stays responsible for how it uses OrviDesk, and this page isn’t legal advice.
July 2024
Lawyers’ use of generative AI: competence, confidentiality, communication, supervision and fees.
In OrviDeskAttorney approval on every matter, disclosure to prospects, and a record of each step.
January 2024
Generative AI in practice: confidentiality, oversight of the tool, fees, and disclosure when a chatbot talks to prospects.
In OrviDeskLexi discloses she’s an AI, and attorneys review before anything is decided.
February 2026
AI in client intake: disclaimers, lawyer review, prompt conflict checks and limits on what AI may do on its own.
In OrviDeskConflicts checked first, disclaimers up front, and approval before action.
Where we are
OrviDesk runs today on a fictional demo firm, and no real client data is in it. Before that changes:
Book a demo
Walk through the safeguards on a live intake with us, and ask about anything your ethics counsel would.