The vote is narrow, but the signal is clear.
Business in Vancouver reported that BC lawyers narrowly rejected a proposal for an AI certification program, and the Law Society of BC's published annual general meeting results show just how close it was: Resolution 7 failed with 845 votes in favour and 986 opposed, or 46.1% to 53.9%.
That is not a profession saying AI has no place in legal work. It is a profession saying that mandatory certification, by itself, is not yet the right control. The stronger lesson is more useful for firms, legal departments, and technology vendors: legal AI apps have to fit the existing duties of competence, confidentiality, supervision, honesty to courts, and client service before any separate AI badge means much.
For Canadian lawyers, the question is not whether AI is allowed. Used carefully, it can support research, summarization, drafting, due diligence, knowledge retrieval, intake, deadline review, and administrative triage. The question is whether the lawyer remains in control of the work and whether the system preserves the obligations that already govern the file.
The merits are real.
Legal AI is strongest when it reduces the mechanical parts of legal work without pretending to replace professional judgment. A well-built legal AI app can search matter materials, identify relevant authorities, summarize long records, draft first-pass memos, compare contract language, surface limitation dates, and help lawyers reuse firm knowledge consistently.
For clients, that can mean faster turnaround, lower cost, and better visibility into the work. For firms, it can mean fewer missed facts, less duplicated research, better internal knowledge capture, and more time spent on strategy and advice. For access to justice, carefully bounded legal AI can help triage problems, explain process, prepare checklists, and connect people with the right human help earlier.
The merit is not magic. It is workflow discipline. The winning systems will be the ones that keep citations visible, separate facts from generated text, preserve an audit trail, and make review easier than blind reliance.
How this fits BC's legal obligations.
In BC, the Law Society has already warned lawyers that generative AI does not suspend ordinary professional duties. A lawyer using AI still has to protect confidential information, understand the limits of the tool, supervise staff and third-party services, check the work product, and avoid misleading clients, courts, or other parties.
The BC courts have also made AI verification a live practice issue. Courts do not care whether a hallucinated case came from a junior, a search engine, or a language model. Counsel remains responsible for the authorities and facts placed before the court.
That is exactly where Canadian-hosted, matter-aware systems matter. For BC firms, the practical control set is straightforward: keep client data in a controlled Canadian environment, restrict access by matter, log usage, cite sources, require lawyer review, and avoid feeding privileged material into public consumer tools.
The provincial rules are converging, even if the wording differs.
Canada does not have a single provincial law society rulebook, but the pattern is consistent. Ontario's technology guidance emphasizes competence, confidentiality, client communication, supervision, and security. Alberta's generative AI playbook frames AI as a tool that must be understood, verified, and governed inside a lawyer's existing professional obligations. Saskatchewan, Manitoba, Nova Scotia, Quebec, and other jurisdictions have taken the same general posture through practice resources, professional responsibility guidance, court notices, or privacy expectations.
The exact labels differ by province. The operating standard does not. A legal AI app has to help lawyers satisfy duties that already exist:
- Competence: lawyers need enough understanding of the tool to know when it is useful, when it is risky, and when a result requires deeper verification.
- Confidentiality and privilege: prompts, uploads, client facts, embeddings, logs, and outputs must be treated as legal file material, not disposable product telemetry.
- Supervision: lawyers remain responsible for AI-assisted work by staff, contractors, and software workflows.
- Accuracy and candour: citations, quotations, procedural statements, and legal conclusions must be checked against primary sources before use.
- Client communication: firms should decide when AI use is material enough to disclose, get consent, or address in retainers and engagement letters.
- Security and outsourcing: vendor review, data residency, access controls, retention, deletion, and incident response are part of the legal risk analysis.
Certification may come later. Controls have to exist now.
The failed BC resolution may slow a mandatory certification approach, but it does not slow the underlying risk. Firms are already using AI. Staff are already pasting text into tools. Clients are already asking why routine legal work still takes so long.
A certification program can be useful if it teaches real operational judgment. But legal AI governance cannot wait for a province-by-province credentialing system. Firms need internal policies, approved tools, training, matter-level permissions, human review standards, and clear records of what AI touched.
For vendors, the bar is equally direct. A legal AI product for BC and Canada should be built around the law society rule environment from the start: Canadian data residency, no customer-data training by default, source-grounded answers, audit logs, matter-scoped permissions, retention controls, and review workflows that assume a lawyer signs off before anything goes outside the firm.
What a compliant Canadian legal AI app should look like.
The safest architecture is not a general chatbot with a legal prompt. It is a matter intelligence system. It knows which file it is working on, which sources are allowed, which user is asking, what was retrieved, what was generated, and what must be reviewed before use.
That is the standard Anitarian is building toward: sovereign Canadian infrastructure, Canadian legal source coverage, matter-scoped retrieval, firm knowledge controls, visible citations, verification workflows, and auditability. The goal is not to let AI practise law. The goal is to give lawyers better tools while preserving the professional duties that make legal advice trustworthy in the first place.
Sources and further reading: Business in Vancouver on the BC certification vote; Law Society of BC annual general meeting voting results; Law Society of BC generative AI guidance; Law Society of Ontario technology guidance; Law Society of Alberta generative AI playbook.