Private AI for law firms in Victoria

A law firm in British Columbia is a private organisation under the Personal Information Protection Act, and separately bound by the Law Society of British Columbia’s rules of professional conduct on confidentiality and client records. Those are two different obligations that happen to point the same way: client information is held in trust, used only for the purpose it was given, and disclosed only where the law or the client requires it.

Solicitor-client privilege survives the tool. Putting a client file into any third-party system is a decision the lawyer makes and must be able to justify — which is why where the information is stored, and who can reach it, is a professional question here and not just an IT one.

Victoria is a government town with a small, relationship-driven bar. The firms here are more often a dozen lawyers around a harbour-view table than a Bay Street floor, and the person doing the research is usually the person who will argue the file.

That shape changes what AI is worth. The gain is not replacing research muscle nobody has to spare — it is answering a client’s procedural question at four o’clock without pulling a senior associate off a brief.

Stored data — accounts, conversations, memory and files — sits in Canada, in AWS Canada Central (Montréal). The models that write answers run on NVIDIA GPU hardware located in Canada and operated from Canada. That second part is the one most vendors gloss: storage residency is not the same claim as inference residency, and for a Victoria firm the difference is exactly the one a privacy review will ask about.

Digby is built here: the tuning and evaluation lab where the open models are adapted and tested runs in Victoria, British Columbia. That does not make Victoria queries faster than Toronto ones — inference serves every Canadian customer from the same Canadian-located GPU hardware — but it does mean the people who tune the models are a short drive from the firms and clinics using them.

If the requirement is that nothing leaves infrastructure you control, Digby can be self-hosted with Docker Compose against any OpenAI-compatible model server. That is a different trade — you run it — and it is documented rather than implied.

1. 01

   Start with know-how, not matters

   put the firm manual and standard forms in a project and let people ask about their own procedures first.

2. 02

   Then one closed matter

   a file already resolved, so the team can check the answers against the record.

3. 03

   Write the rule down before opening it up

   what may go in, what never does, who reviews an output before it is used.

Victoria firms and organisations often work across provincial lines, and the statute changes at the border. The same law firms material, set up for the next province over: