Thought Leadership
Early court decisions on generative AI: What Canadian businesses need to know
Matthew N. Craig and Sam Schonfeld
Regardless of which industry you are a part of, you have likely noticed the adoption of generative artificial intelligence (“GenAI”) all around you. Increasingly, employees and businesses are using artificial intelligence to review documents, prepare summaries, assist with developing arguments, or draft letters and agreements.
If you use GenAI in your business, it is important to understand how early decisions from courts and tribunals in the United States and the United Kingdom may influence developments in Canadian law – particularly in relation to issues of privilege and confidentiality.
United States
In United States v Heppner, a judge of the United States District Court for the Southern District of New York held that documents generated by a criminal defendant who used Claude, a popular GenAI model built by the company Anthropic, were not privileged or protected by the work product doctrine.
Bradley Heppner, the defendant and a former CEO of GWG Holdings Inc., used GenAI to generate 31 documents to formulate potential defences to the fraud charges against him. He then presented these documents to his attorneys.
The Court found that solicitor-client privilege attaches to communications that are:
- between clients and their attorney;
- intended to be kept confidential, and are kept confidential; and
- occurring for the purpose of obtaining or providing legal advice.
The Court held that Heppner’s use of GenAI is not protected by attorney-client privilege, finding that the claim of attorney-client privilege failed the first two, if not all three, elements of the test.
On the first and third points, the Court characterized Heppner’s chat history with the GenAI tool Claude as “the discussion of legal issues between two non-attorneys”. The judge concluded that Heppner communicated with Claude of his own volition and not at counsel’s direction, and also found that Claude expressly disclaimed providing legal advice (“I’m not a lawyer and can’t provide formal legal advice or recommendations”).
The Court’s finding on the second point could have implications beyond claims of solicitor-client privilege. Specifically, the judge found that Anthropic’s privacy policy explicitly stated that Claude collected data on user inputs and outputs that the company could later sell or give to third parties, and therefore communications between Heppner and Claude could not be intended to be kept confidential.
In Warner v Gilbarco, Inc., a judge of the United States District Court for the Eastern District of Michigan reached the opposite conclusion.
In the Warner case, a self-represented plaintiff used GenAI to assist her with drafting litigation materials in relation to a wrongful dismissal case against her employer.
The defendants requested production of all documents related to the plaintiff’s use of third-party AI tools. The Court rejected the request for production, finding that the documents were entitled to protection under the “work product doctrine”. The case was decided based on a different legal principle, but the Court notably held that Claude, ChatGPT, Gemini, and other GenAI models “are tools, not persons”, finding that GenAI merely helped the plaintiff express legal opinions which were only in her mind. While discussion with another person may have led to a waiver of privilege, use of a tool did not.
In Morgan v V2X, the United States District Court for the District of Colorado considered the use of GenAI in context of an employment discrimination case. In that case, the defendant employer sought to compel the self-represented plaintiff to disclose the name of the GenAI tool used in connection with the case.
The self-represented plaintiff resisted disclosure of the name of his chosen AI, arguing that his choice of AI tool fell within the work product doctrine, as it reflected his mental impressions and litigation strategy. After considering Heppner and Warner, the Court noted that sharing information with an AI platform does not automatically waive work product protection, since AI tools are not adversaries and disclosure to them does not substantially increase the likelihood that an adversary will obtain the materials. However, the Court ultimately ordered the plaintiff to disclose the name of his AI tool on the basis that the plaintiff had failed to demonstrate that doing so would reveal his mental impressions or strategy. The Court also held that the defendant had a legitimate interest in knowing which system had received its confidential information.
United Kingdom
In UK v Secretary of State for the Home Department, the Upper Tribunal (Immigration and Asylum Chamber) considered a slightly different scenario – a lawyer’s use of ChatGPT. The Tribunal found that putting emails to clients and other case information into ChatGPT amounts to placing the information on the internet in a public domain. Accordingly, the Tribunal found that the use of ChatGPT resulted in a breach of client confidentiality and a waiver of legal privilege.
Of note, the Tribunal distinguished open-source AI tools from closed-source AI tools, noting that AI tools which do not place information in the public domain are available for use without the same confidentiality risks.
What this Means in Canada
There are some differences between the legal principles considered in these decisions and those that are likely to apply in Canada. Nonetheless, when faced with similar fact scenarios and questions, the reasoning in these decisions could be considered by Canadian courts and tribunals.
Canadian businesses should proceed with caution when using GenAI to review or produce information that is intended to be privileged or confidential.
Until the law develops in Canada, uploading sensitive or privileged information into GenAI tools risks a loss of confidentiality and/or waiver of privilege, which could result in an order that the information be shared with adverse parties or competitors.
This client update is provided for general information only and does not constitute legal advice. If you have any questions about the above, please contact the authors.
Click here to subscribe to Stewart McKelvey Thought Leadership.