Factor
Sensitivity
Intake is where a client says the most and knows the least about who is listening. CounselRoom treats that conversation as matter content from the first message, not as a lead in a marketing tool.
Professional obligations
The rule asks for reasonable efforts. This is the record of yours.
No software can promise you privilege. What software can do is help you run a process you would be comfortable describing, and leave proof that you ran it.
In re: the reasonable-efforts standard
-
Rule 1.6(c) asks a lawyer to make reasonable efforts to prevent the unauthorized disclosure of, or unauthorized access to, information relating to a representation.
Model Rule 1.6(c)
-
Comment [18] sets out how that is judged: how sensitive the information is, how likely disclosure becomes without additional safeguards, what those safeguards cost, how difficult they are to use, and whether they would get in the way of actually representing the client.
Model Rule 1.6, cmt. [18]
- Read those together and one thing follows. The rule never asks you to guarantee an outcome. It asks you to run a defensible process and to be able to show it.
- So a vendor promising you privilege is promising something the rule does not recognize and no software can deliver. Privilege is decided later, by a court, on the facts, under the law of that forum. What is decided by you is what you did.
- We do not sell you a guarantee. We give you the record of what you did: who had access and why, what the AI produced, what a lawyer approved, when, and what changed afterward. That is the artifact the standard is actually asking for.
02 · Comment [18]
How reasonableness gets judged.
Four of the five factors are things a tool can move. The fourth is the one most vendors will not mention.
Factor
Likelihood of disclosure
The realistic alternative is a consumer chatbot, a text thread, or a personal inbox, none of which your firm controls. Moving the conversation into a room you open and close changes that exposure directly.
Factor
Cost and difficulty
Honestly: this is software your firm pays for, and an invited room is more friction than answering a text. The rule expects you to weigh that against how sensitive the matter is. Sometimes a phone call is the reasonable answer.
Factor
Getting in the way
A safeguard nobody uses protects nobody. Your client answers questions in a chat, on their own schedule, which is why intake actually gets finished rather than abandoned halfway.
03 · The map
Obligation by obligation.
Your duties, and what you would have to show for each one.
| Obligation | What the rule asks | What you have to show for it |
|---|---|---|
| Rule 1.6(c) and Comment [18] | Reasonable efforts against unauthorized disclosure of, or access to, client information. | Access is checked on every request against who you are, your firm, the specific matter, and your role. Invitations are single use. A firm administrator can see that a matter exists without being able to read it. Attorney private notes and AI payloads are encrypted where they are stored. |
| Rule 1.1 and Comment [8] | Competence, including keeping up with the benefits and risks of the technology you use. | Every AI run records which model answered, which version of the prompt asked, and a fingerprint of what went in and came out. You can state what ran, on what, and when, rather than describing the tool in general terms. |
| Rule 5.1 | Supervisory lawyers make reasonable efforts to ensure the firm conforms to its professional obligations. | A responsible attorney is on the matter before the client can enter. Any attorney or staff member can pause the AI, and the firm has a switch that stops it everywhere. Reviews are recorded against the person who made them. |
| Rule 5.3 | Responsibility for nonlawyer assistance, which is the frame ABA Formal Opinion 512 uses for generative AI. | AI output is provisional by default and never reaches your client as approved until an attorney approves it. It is checked against a schema before it is stored, and it identifies itself to your client as an AI rather than a member of the firm. |
| ABA Formal Opinion 512 (2024) | Competence, confidentiality, communication, and supervision when using generative AI, including diligence on what the vendor does with your data. | Your client is told up front that they are talking to an AI. Notifications and logs carry no matter content. Exports label what the AI produced separately from what an attorney reviewed. Provider terms are part of pilot onboarding, in writing, rather than something you have to infer. |
This maps obligations to product behavior. It is not a legal opinion, it is not legal advice, and no bar association has reviewed or approved this product or any other. Rules and their numbering vary by jurisdiction; the citations here are to the ABA Model Rules.
Noticeread this part too
What this page is not.
- It is not legal advice, and reading it does not create an attorney-client relationship with anyone.
- It is not a compliance certification, and it is not a bar association's endorsement. No bar has approved this product, or any product.
- It is not a substitute for your own judgment. You decide whether any tool meets your obligations, in your jurisdiction, on this matter.
- It is not settled ground. This area of the law is new and moving, and we will keep this page current as it moves.
Reviewed by the responsible attorney
Run a process you would be glad to describe.
CounselRoom is in an invite-only pilot for small and midsize firms.