Artificial Intelligence Hallucinates Case Law Introduced In a Canadian Court

a lawyer sits behind a scale balancing a rendering of AI and a brain

We knew it was bound to happen. And, now it has. The misuse of artificial intelligence (AI) has found its way into Canadian courts - and it was an officer of the court who was ultimately responsible for bringing it there.

On January 23, 2024, Global News reported that British Columbia lawyers Lorne and Fraser MacLean had discovered fake case law entered into court by opposing counsel for a civil case involving a “high-net-worth family matter, with the best interests of children at stake.”

Lawyer Chong Ke told the court she did not intentionally try to mislead the court by submitting legal briefs with this fake case law; rather, the AI tool (ChatGPT) she allegedly employed to assist in her research had a “hallucination” that prompted it to create realistic sounding fake information.

In this blog post, I’ll outline how this case makes BC ground zero for fake AI cases in Canada, explain why it was human error rather than a glitch in AI software itself that lead to this very concerning situation, and suggest why incidents such as this one reaffirm my belief that we should not shy away from this exciting new technology, but rather regulate it and refine it so that it can help lawyers do their work as opposed to throwing our profession into disrepute.

What Is AI and How Was It (Mis)Used in BC?

AI trains machines through experiential learning to think and act like humans. AI systems adjust to new data inputs to alter and refine their outputs similar to how humans adjust their thinking based upon learning from their lived experience or being taught new information.

Well-known AI systems currently available derive their knowledge from deep learning and natural language processing. For example, ChatGPT is a large model natural language processing chatbot that permits users to engage in a conversation. ChatGPT analyzes prompts and replies to provide context for the discussion. Through experimentation with the tool, users have discovered this program’s versatility makes it suitable for some creative endeavours and for writing and/or correcting computer code.

Despite these exciting possibilities, ChatGPT’s developer OpenAI notes it has some serious limitations. On its website, OpenAI explains: “ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this issue is challenging, as: (1) during RL training, there’s currently no source of truth; (2) training the model to be more cautious causes it to decline questions that it can answer correctly; and (3) supervised training misleads the model because the ideal answer depends on what the model knows, rather than what the human demonstrator knows.”

In several high profile cases in the United States, the United Kingdom, and now Canada, the chatbot’s “plausible-sounding but incorrect” answers or “hallucinations” found their way into briefs submitted to courts. Unless ChatGPT is fact-checked, there is potential for this predictive text tool to make an incorrect prediction that could be subsequently presented as fact.

Ke’s statements to the court in response to the discovery of the AI-produced fake case submission suggests her actions were accidental and based on ignorance of the technology rather than evidence of an attempt to intentionally mislead the court.

Nevertheless, whether intentional or not, the damage done to the reputation of the court and our profession are very real - particularly because this case involved the interests of children.

While reasonable people can come to different conclusions about the facts of a matter before a court, it is incumbent on our legal system to ensure evidence heard before a court is real and accurate.

Lawyers, as officers of the court, have a duty to be forthright and truthful; they can be disciplined if they are not being honest with evidence they are presenting. In Ke’s case, the Law Society of British Columbia, which issued a warning and guidance to lawyers about AI use in late 2023, could investigate the matter and take disciplinary action.

Convincing to a Fault.

When ChatGPT is prompted to write a convincing legal brief on a topic, it draws on what it knows about the form and style of legal briefs and publicly accessible data on a topic to generate a response. But what if it can’t identify an existing case to help it make a convincing and persuasive argument? Why not draw on aggregate data to create a case that would help it make its point?

ChatGPT worked according to design; but what it was designed to do is not appropriate for a court of law. Courts weigh the value of verifiable facts to determine a truth that is, depending on the type of case, either beyond reasonable doubt or more likely than not based on the balance of probabilities; ChatGPT’s programming employs “truthiness” to generate text it predicts that a user wants based on context.

Beyond the inherent ethical questions that arise from professionals using AI to create a product without acknowledging the source, the emergence of fake cases in court could sully jurisprudence if judges are not careful to fact check case law presented in briefs. What occurred in this case is an enormous waste of court resources and it’s rightly sending shockwaves across the country.

The widespread availability of AI tools could also have profound effects on other types of evidence introduced in matters before the courts. How will courts respond to submissions in small claims courts and tribunals where individuals may be self-represented and lack the oversight of professional regulatory bodies that will hopefully deter this practice from becoming commonplace? As these tools improve to a point where they can be employed to forge evidence that humans cannot identify as fake, how will our justice system respond?

Thankfully, our governments and institutions are tackling these issues head on by demanding guarantees from AI developers, and developing regulations to protect our society from misuse of this technology. For example, the Federal Court’s Strategic Plan (2020-2025) noted its interest in this emerging field and it has issued interim principles and guidelines and notices in response to developing events.

Many provinces have recently amended their Rules of Civil Procedure in response to the potential use of AI. Ontario, for example, now requires lawyers to certify “the authenticity of every authority” listed in their factums.

Embrace New Technology, Responsibly.

Long-time readers of this blog will know that Gluckstein Lawyers, our team of personal injury lawyers and team members prides itself on being an early adopter of technology, including generative AI. Technological advances have created products that can be transformative in a practice such as ours. The cost savings and increased efficiency we’ve found by employing new tools judiciously has allowed us to free up staff time and redirect it to better serve our clients.

When it comes to integrating AI into our operations, clearly I’m not a Luddite in the way we’ve popularly come to understand the term.

But, perhaps I do share some affinity to the historical Luddites in terms of their actual concerns about using new technology. As Kevin Binfield, editor of Writings of the Luddites, notes in a Smithsonian Magazine article, the Luddites “were totally fine with machines,” but opposed manufacturers using them in “a fraudulent and deceitful manner” to circumvent standard labour practices.

He explains: “They just wanted machines that made high-quality goods, and they wanted these machines to be run by workers who had gone through an apprenticeship and got paid decent wages. Those were their only concerns.”

If AI, for all its potential and limitations, is respected by its users, it can be a net benefit in a variety of sectors, including legal practice. But, if this technology is not well regulated and used responsibly, we run the risk of experiencing more incidents like the BC case.

Artificial intelligence tools such as ChatGPT are groundbreaking technological advancements. Like any disruptive technologies, they have the potential to both harm and help humanity. Ultimately, it is up to humanity to find ways to employ this technology for the benefit of humankind by refining it and regulating it to limit the possibility of unintentional or malicious misuse.

Share

Subscribe to our Newsletter

Sign me up