Articles
<< back to all ArticlesThe Privilege of AI – Another Risk for Litigants
Whether we like it or not, the legal system’s relationship with artificial intelligence is rapidly changing. There are ways to use AI safely. There are also ways to use AI dangerously. Many are now aware of AI hallucinations (fictitious cases the AI agent treats as real) and the consequences from relying on such imaginary authorities. In Reddy v Saroya, for example, the Alberta Court of Appeal ordered a lawyer to personally pay $17,550 after relying on a contractor’s written submissions containing fictitious cases. Many justice system participants are now aware of the dangers of relying on AI agents and the need to review for AI hallucinations.
One newer risk facing litigants is the interaction of AI agents and privilege. Solicitor-client privilege is a foundational component of Canada’s legal system. When someone seeks or receives confidential legal advice from a lawyer, that communication is “privileged”. At a high level, this means that the conversation is private. Other than in very narrow exceptions, the lawyer is bound to keep confidential what they discussed, and the client cannot be forced to reveal it. Solicitor-client privilege arises if three conditions are met:
- the communication is between a lawyer and their client;
- the communication entails the giving or seeking of legal advice; and
- the parties intended the communication to be confidential.
Solicitor-client privilege facilitates “full and frank communication in the seeking and giving of legal advice”. It encourages clients to speak openly and candidly with their lawyer, which enables lawyers to provide the best legal advice they can to their clients. Beyond helping individual clients, solicitor-client privilege is a necessary component of a well-functioning legal system.
There is a risk that conversations with AI agents may not be protected by solicitor-client privilege. Where a client communicates with an AI agent, each of the three elements above may be missing. First, AI agents may not be “lawyers”. Second, communications with AI agents may not involve the giving or seeking of legal advice. After all, many AI agents profess to be unable to give legal advice. Third, most public AI agents are expressly not confidential: when you sign up to use one, you often agree to Terms of Service which allow the company operating the agent to use your inputs and the agent’s outputs for training the AI agent.
While there do not appear to be any Canadian court decisions discussing this problem yet, this is likely only a matter of time. Courts in other jurisdictions have begun addressing the issue, generally warning that conversations with public AI agents may not be protected by solicitor-client privilege. In the United States, a Court in the Southern District of New York found that a client’s communications with Anthropic’s “Claude” AI agent were not protected by attorney-client privilege (the United States’ equivalent to solicitor-client privilege), because Claude is not a lawyer and the conversation was not confidential.
In the United Kingdom, the Upper Tribunal issued a warning after a lawyer uploaded client information into ChatGPT—a publicly available AI agent: “to put client letters and decision letters from the Home Office into an open source AI tool, such as ChatGPT, is to place this information on the internet in the public domain, and thus to breach client confidentiality and waive legal privilege”.
In Australia, while not addressing the issue head on, a Court warned that inputting documents “into a generative AI program which stores, collates and replicates data may waive privilege”.
It is only a matter of time before we see Canadian courts grappling with privilege and AI agents. Absent authoritative guidance from the Court, lawyers and litigants should exercise caution in putting confidential information into publicly available AI tools. These “conversations” may not be given the protections granted to communications with lawyers. Similar concerns arise for clients who have lawyers. Putting the advice received from a lawyer into an AI agent may inadvertently waive privilege and risk exposing that communication to others.
This post is meant to provide information only and is not intended to provide legal advice. Although every effort has been made to provide current and accurate information, changes to the law may cause the information in this post to be outdated.
