Can Alberta law firms use ChatGPT without breaking client confidentiality?
What the Law Society of Alberta's Generative AI Playbook and model policy actually say, and how a small firm can follow them.
The short answer
The Law Society of Alberta doesn't ban generative AI. Its Generative AI Playbook (last updated February 2026) explains how lawyers can use it, and its top recommendations include protecting client confidentiality, informing clients and supervising staff (Playbook). On confidentiality it is blunt: "Never include confidential or potentially identifying information in prompts" and "Use only non-identifiable information in prompts." The Playbook ties this to the duty of confidentiality in Rule 3.3 of the Code of Conduct.
So the question for a small firm isn't whether to use AI. It's how to make sure nobody pastes a client's name, file number or facts that identify them.
What the Law Society's resources say
- Confidentiality. The Playbook warns that tools like ChatGPT "were not built with confidentiality in mind" and that uploaded content may be used to train the tool and can't always be retrieved (Playbook).
- Platform terms. Free versions of several major AI platforms use user data for training, and even enterprise versions may access content to enforce their policies (Terms and Conditions of Gen AI Platforms).
- A written firm policy. The Law Society publishes a Model Generative AI Use Policy for firms to adapt, and says AI use policies also speak to lawyers' duty to supervise articling students and staff (Why a Generative AI Use Policy?).
- Staff guide. The companion guide for lawyers and staff says "you should never input any confidential, personal or private information into a Gen AI prompt, including in documents you may upload" (guide, PDF).
- Tell clients. The Playbook recommends explaining your AI use in your retainer letter.
Where de-identification helps, and where it doesn't
A de-identification tool replaces names, file numbers, addresses, phone numbers and account numbers with tags before the text reaches the AI, then restores them in the answer. That directly supports the "no identifying information in prompts" rule.
It doesn't make the remaining facts non-confidential. A dispute between "#PERSON_1" and "#ORG_1" over a specific Edmonton development may still identify the client. The Law Society's wording covers confidential information, not only names. A lawyer still decides what may be shared at all, and some matters shouldn't go near a consumer AI tool even after masking.
A practical setup for a small Alberta firm
- Adopt a policy. Start from the Law Society's model policy. For support staff who need something shorter, our one-page staff AI-use policy covers the basics in plain language.
- Approve specific tools. Read the terms of each tool you allow. Prefer paid business plans that don't train on your data.
- Put a safeguard in the browser. Install a masking step on every computer and lock the settings so it can't be switched off casually.
- Train and supervise. Fifteen minutes with a real example. Treat AI output as a first draft and verify every citation.
- Tell clients. Add a sentence on AI use to your retainer letter.
- Keep a record. Hushtag's AI-use record shows when details were hidden and on which AI site, without storing any client text.
Sources
- Law Society of Alberta: The Generative AI Playbook (last updated February 2026)
- Law Society of Alberta: Why a Generative AI Use Policy? (Model Generative AI Use Policy)
- Law Society of Alberta: How to Use Generative AI in Your Legal Practice, A Guide for Lawyers and Staff (PDF)
- Law Society of Alberta: Terms and Conditions of Gen AI Platforms
Links checked October 18, 2026. Laws and guidance change; rely on the official versions.