Eleven courts, nine law societies and a growing list of other bodies have all said the same first thing: generative AI can be used. What differs, and matters more, is what each of them attaches to that yes.
Key takeaways
Yes — every Canadian body that has weighed in agrees on that much. The Law Society of Alberta’s own survey of the field puts a number on how many bodies that is: “since ChatGPT entered the legal lexicon, eleven Canadian courts, nine law societies, two professional liability insurers, one provincial government, the Canadian Judicial Council, Canadian Bar Association and the College of Patent Agents and Trademark Agents have all issued guidance documents regarding the use of Generative AI” — and, its opening line continues, “none have attempted to ban its use.” (LSA, Gen AI Rules of Engagement for Canadian Lawyers) The real question is not whether, but under what conditions.
The Law Society of British Columbia’s practice resource states the confidentiality problem plainly: “consider carefully what information needs to be supplied to the generative AI tool with a particular attention to client confidential information. Ideally, client confidential information, including any information identifying the client, would be omitted from anything that was supplied to the generative AI tool to maintain client confidentiality.” (LSBC, Guidance on Professional Responsibility and Generative AI) The reason is not abstract: “as generative AI tools (especially those powered by LLMs) rely on previously inputted information to learn how to produce more accurate answers in the future, the generative AI tool may reuse the confidential information you supply for other purposes over which you may have no control.” (LSBC, Guidance on Professional Responsibility and Generative AI) That is a mechanical claim about how the tool works, not a generic caution — it is the reason redaction, not just discretion, is the recommended default.
The same LSBC guidance ties competence to the Code of Professional Conduct directly: “before using generative AI in your client work, it is prudent to seek out more information and training on how to use the tool,” because Code rule 3.1-2 “requires that lawyers must perform all legal services undertaken on a client’s behalf to the standard of a competent lawyer,” with rule 3.1-1 defining what that standard covers. (LSBC, Guidance on Professional Responsibility and Generative AI) Competence and confidentiality are linked in practice: a lawyer who does not understand how a given tool retains or reuses input cannot make an informed judgment about what is safe to type into it.
Two law societies converge on the same billing principle from different directions. Manitoba, in its April 2024 guidelines, put it as a windfall rule: “fee arrangements should not generate an inappropriate windfall for a lawyer arising from the efficiencies created by using an AI tool to perform a certain task” — while making clear it remains “appropriate to charge for the time spent in crafting and refining AI inputs and prompts and in reviewing, confirming, analyzing and editing generative AI tool output.” Ontario’s Law Society Futures Committee White Paper reaches the same place through hourly billing specifically: “if a licensee is billing by the hour, they can only charge for the time actually spent by the licensee on the file, even if a generative AI tool has made the task much more efficient.” (LSA survey, citing Manitoba and LSO guidance)
If AI-generated content ends up in a document filed with the Federal Court, a separate question attaches — not billing or confidentiality, but disclosure to the court itself. The Court’s own test is whether the AI’s role “resembles that of a co-author”; merely using AI to “suggest changes, provide recommendations, or critique content already created by a human” does not require a Declaration, but content “directly provided by AI” does. (Federal Court, Notice to the Parties and the Profession, May 7 2024) The mechanics of that test are covered in full on this hub’s piece on the Federal Court’s filing rules; the point here is that it is a distinct obligation layered on top of confidentiality, competence and billing, not a replacement for any of them.
The Law Society of Ontario’s Futures Committee White Paper “did not insist that lawyers inform their clients every time they use Gen AI,” and instead set out four questions to consider: “will the use of Gen AI necessarily be disclosed publicly (for example if Gen AI is being used in preparation of a court document before a court that requires such disclosure)?”; “does the client reasonably expect that the material being prepared by Gen AI would actually be prepared by a licensee?”; “are there reputational or other forms of risk to the client that could arise from the use of Gen AI?”; and “does use of Gen AI require inputting of the client’s personal or proprietary information?” (LSA survey, citing LSO Futures Committee White Paper) Where any of those point toward disclosure, the White Paper says lawyers “should be prepared to explain to their clients how they use the technology in their matter, any associated risks, and what steps they take to mitigate same.”
A sole practitioner in Vancouver is drafting a demand letter and wants an AI tool to tighten the language. Confidentiality first: the settlement figures and the client’s name should ideally never reach the tool at all — a redacted version of the draft, with placeholders for the identifying details, does the same drafting job without the exposure the LSBC guidance describes. Competence next: before relying on the output, the lawyer needs enough understanding of the tool to know whether it might soften a demand’s legal effect or introduce a concession that was not intended — that is what Code rule 3.1-2’s competence standard actually requires here. Billing: if the AI pass took five minutes and saved what would otherwise have been forty-five, only the five minutes plus review time goes on the invoice, under both the Manitoba and Ontario positions. And because no court filing is involved yet, the Federal Court’s Declaration test does not apply — it would if the same letter later became an exhibit drafted with undisclosed AI assistance in a factum.
The confidentiality, competence and billing principles above are consistent everywhere they have been stated, because they restate existing professional-conduct rules that already applied before generative AI existed — a lawyer always had a duty to protect client information and bill only for time worked. What genuinely differs by province is the court-facing layer covered elsewhere on this hub: whether a specific practice direction requires AI use to be disclosed to the court, and what a fabricated citation triggers once it is filed. Alberta, Manitoba, Yukon, the Federal Court and Ontario each answer that piece differently, which is why the professional-conduct answer above (“yes, under conditions”) does not by itself tell a lawyer what a specific court will require.
Related: why professional bodies are cautious on AI, the Federal Court’s rules on AI in filings, and regulated professions and AI guidance
The Law Society of BC's guidance treats this as the central risk: ideally, client confidential information would be omitted from anything supplied to the tool, because a generative AI tool may reuse what you input to improve future answers, in ways you cannot control.
No, according to both Manitoba's and Ontario's guidance. You can bill for the time actually spent crafting prompts and reviewing, confirming and editing the output — not for the time the task would have taken without the tool.
Not automatically. The Law Society of Ontario's four-factor test looks at whether the use will be publicly disclosed anyway, whether the client expects a licensee's own work, reputational risk, and whether the client's own confidential information was input — not a blanket rule.
Confidentiality, competence and accountability apply to any regulated professional evaluating a vendor, not only lawyers.
