LSO AI Policy Template: A Generative AI Policy for Ontario Law Firms (2026)

The Law Society of Ontario (LSO) has not published an AI policy template, and it is unlikely to. What it published, in April 2024, is a white paper, 'Licensee Use of Generative Artificial Intelligence', and a practice resource, 'Generative AI: Your Professional Obligations', which explain how the existing Rules of Professional Conduct apply when a lawyer or paralegal uses generative AI. A firm policy is how you convert those obligations into rules your people can follow. This guide covers what the LSO said and what the policy needs to contain.

The six obligations the LSO tied to generative AI

• Competence (Rule 3.1-2): licensees must understand the technology they use and its risks. You cannot competently use a tool whose failure modes you do not know, and hallucinated authorities are the failure mode the courts have already punished.
• Confidentiality (Rule 3.3): the duty extends to information entered into a generative AI platform. A prompt containing client information is a disclosure to the vendor, so the tool's data handling determines whether the duty is met.
• Honesty and candour: clients must be told where AI use may affect their interests, the outcome, or the cost of the service.
• Supervision: professional judgment cannot be delegated to technology. AI output is reviewed the way a junior's draft is reviewed, and the licensee remains responsible for it.
• Reasonable fees and disbursements: time saved by AI cannot be billed as if it were spent, and AI costs passed on to clients must be disclosed and reasonable.
• Not misleading the tribunal: every citation and authority produced with AI assistance is verified before filing, and any court-specific AI disclosure requirement is followed.

The risks the white paper names

The LSO's white paper is candid about where generative AI goes wrong in practice: the unanticipated spread of confidential information, hallucinations and inaccurate output, bias in results, damage to the licensee-client relationship, and tools that end up effectively giving legal advice. A useful firm policy addresses each of these with a rule, not a warning.

Courts have their own requirements

Separately from the Law Society, some courts now regulate AI in filings. The Federal Court's practice notice on the use of AI in court proceedings requires a declaration when AI-generated content is included in documents submitted to the Court, and several Canadian courts have issued similar directions. Your policy should require lawyers to check the rules of the court they are before and to verify every authority personally, since the sanctions in the reported cases fell on the lawyer, not the tool.

What an Ontario law firm AI policy must contain

• Scope: lawyers, paralegals, articling and law students, clerks, and administrative staff; every generative AI tool including features embedded in research platforms, document management, and email.
• Approved tools by data category: which tools may be used with client-identifiable information (only those with enterprise terms, no training on inputs, and appropriate data residency), which may be used only with anonymized or public material, and which are prohibited outright.
• Confidentiality rules: no client information, privileged material, or file-identifying detail in any tool not approved for it; personal accounts prohibited for firm work.
• Verification: all AI-assisted research, citations, and drafting are verified by the responsible licensee before use; no filing contains an unverified authority.
• Client communication: when clients are told about AI use, how consent is recorded where it matters, and how AI use is reflected in retainer agreements.
• Billing: AI use does not inflate time entries; any AI-related disbursement is disclosed.
• Court compliance: the licensee checks and follows the AI directions of every court and tribunal before filing.
• Supervision and training: who approves new tools, how staff are trained, and how acknowledgments are collected.
• Incident reporting: what happens when confidential information reaches an unapproved tool or an AI error reaches a client or a court, with self-reporting encouraged.

Small firms and sole practitioners

The LSO's obligations do not scale down for firm size, but the policy can. A sole practitioner's AI policy is one or two pages: the tools you use, what never goes into them, that you verify everything, that clients are told when it matters, and that you follow court directions. Having it written down is what allows you to show, if it is ever questioned, that you understood the technology and used it deliberately.

The fast path

Our generator produces a law firm AI policy with privilege and confidentiality clauses, mandatory verification of AI-generated authorities, Canadian privacy provisions, and Canadian spelling when you select Canada as your jurisdiction. Preview the full document free before paying. It is a self-help template rather than legal advice, and it does not replace your own reading of the LSO's guidance.