Confidentiality · Model Rule 1.6

Confidential legal AI, measured against Rule 1.6.

Rule 1.6(c) asks for reasonable efforts to prevent unauthorised disclosure. Not transmitting the material anywhere is the strongest version of that effort.

Request a Briefing See what it does

How the rule maps onto the architecture

Each obligation, and the thing in the product that answers it.

Reasonable efforts, 1.6(c)

Client information is read on firm-owned hardware with no outbound path. There is no transmission to safeguard because there is no transmission.

Competence, 1.1 comment 8

Understanding the benefits and risks of the technology is easier when the technology is a machine in your office rather than a service you cannot inspect.

Supervision, 5.1 and 5.3

Output goes to a named reviewer before it is relied on. The review decision is recorded, so supervision is evidenced rather than assumed.

Client instructions

A matter where the client has restricted automated processing can be recorded as such and enforced, rather than depending on someone remembering.

What stays true on every matter

These do not change by practice area, firm size, or the question asked.

Nothing leaves the firmNo cloud, no API call, no telemetry. The matter is read on hardware the firm owns.

Owned, not subscribedA one-time purchase. No per-seat monthly fee and no usage meter.

Citations verified on-machineAuthority is checked against a local corpus, not recalled from a model’s memory.

The entire matterFull workup across the whole file — not a chat window over one document.

Runs on your workstationA current RTX-class GPU. Designed to run with no internet connection at all.

Take the architecture to your ethics counsel.

It is short enough to explain in a paragraph and specific enough to be assessed. We will put it in writing for you.

Request a Briefing

Questions firms ask

The answers below are specific to this page’s subject.

Can you certify that we are Rule 1.6 compliant?

No, and nobody honestly can. Compliance is assessed against a firm's whole practice, not a single tool. What we can say precisely is that client information is not transmitted to any third party by this product.

Do we need client consent to use it?

That is a judgment for the responsible lawyer and depends on the engagement. The architecture removes the third-party disclosure that usually drives the question.

What about state bar guidance that differs from the ABA?

Several state bars have issued their own AI guidance and the common thread is confidentiality, competence, supervision and candour. An architecture where the data never leaves the firm sits well against all of them.

Is there an audit trail?

Yes. What was run, when, by whom, and what a reviewer decided about the output — held locally, like everything else.

Other pages covering the same ground from a different angle.

Secure on-premise AI for attorney-client privilege Private AI for privileged documents Zero data leave / no cloud legal AI Secure legal document analysis AI Air-gapped AI for law firm data security Stop sending privileged data to the cloud

See it run on a real matter.

A briefing walks the whole path — intake, workup, the Adjudicator's pass, and the work product that comes out the other side.