AI conjures ‘hallucinated cases’ in Calgary lawyer’s closing arguments

A Calgary lawyer is facing a Law Society of Alberta hearing after AI-generated closing arguments cited two cases that do not exist.

A Calgary lawyer is facing a Law Society of Alberta hearing after misusing artificial intelligence (AI) in composing her closing arguments in a recent court case.

Folasade Olugbemi with Calgary law firm EO Law was representing a mother in a child custody battle in the Alberta Court of Justice.

While she was successful in having a permanent guardianship order dismissed and her client’s children returned to her on a six-month supervision order, the justice in the case learned AI was used for her closing arguments in the case, her statement referred to “hallucinated cases.”

“Two of the three cases cited in the written closing submissions do not exist. The mother’s counsel stated that she is not relying on the impugned cases, which goes without saying given that the impugned cases do not exist,” Justice S.H. Borsic said in her reasons.

• Sign up for breaking news alerts from CTV News, right at your fingertips
• The information you need to know, sent directly to you: Download the CTV News App

According to the court, AI can be used by legal counsel, but efforts have been made to ensure “caution and care” is taken when relying on AI in court submissions.

Those steps include a formal notice signed by the chief justices of all three courts in Alberta in 2023, which “is readily available on the websites of all three courts.”

Doug King, a professor of justice studies at Calgary’s Mount Royal University says AI, or large language models, is being used in almost every profession.

“There are actually four or five proprietary AI applications designed specifically for law. They’re used by large law firms and small law firms,” King said.

“But you still have to have a human eye on the end result. That didn’t happen in this particular case.”

AI can have benefits for the legal system, King said, by helping organize complex cases and even come up with a draft, but that’s where it should stop.

“That’s where the eyes have to be put on it,” he said.

Lawyer’s responsibility to fact-check

Olugbemi did apologize to the court and other counsel in the case and admitted she relied on her assistant’s help in preparing her written closing submissions because of “an emergency,” Borsic said.

“She acknowledged that it is her responsibility to oversee and review the work that she assigns to her assistant. She also advised that she spoke to a practice advisor at the Law Society of Alberta,” she wrote.

“However, this does not absolve the mother’s counsel of responsibility. It was incumbent upon her to strictly adhere to the notice and to ensure that those to whom she assigns work do the same.”

In the case, King suggested the counsel’s assistant might not have been well trained in the use of AI and might have also not been using one of the firm’s proprietary tools.

“They might have been using something like ChatGPT or Grok,” he said.

“The way these large language models work is it’s trying to find information. It’s basically trying to summarize existing information it can find.

“You know you’ll get information that’s not accurate.”

No matter what happened, King says it’s the lawyer’s responsibility to read over the final document before it was submitted.

“It was still incumbent upon the lawyer to read the final document and any lawyer who’s reading the final case summary and is encountering cases that are named – they better know those cases.”

Could be much more AI misuse: King

The justice said this wasn’t the first time this has happened in an Alberta court.

“Regrettably, I am not the first member of the judiciary in this province who has had to address the misuse of artificial intelligence in court submissions,” Borsic said.

King isn’t sure how many Canadian cases have been labelled as misusing the technology, but warns that there could be many more.

“Here’s the frightening thing – they’re not all getting caught," he said.

“AI and the use of AI is going to make it into all professions and no one’s going to be able to recognize it.

“It’s the same old adage – the time you catch someone doing something wrong isn’t the first time they’ve done it. It’s just the first time you caught them.”

King calls the present situation with AI “a transition phase.”

“We’re in that strange transition of, ‘it’s a great tool, I want to use it, but am I using it ethically and is it accurate?’”

As time passes, the technology will improve in its accuracy, so King says its up to everyone to use it ethically.

“You always need to know when you’re before a judge in a court of law that your evidence, your testimony, your statements have to be both credible, which means you’re not lying, and have to be reliable in the sense that you’re accurately recalling things even though you’re telling the truth,” he said.

“You have challenges here that if you use AI knowingly and you submit something that you know some of these cases are hallucinations and you know that and you submit, that’s an issue of credibility.”

At the outset, Borsic called Olugbemi’s infraction something “serious” that “must be addressed.”

“I do not know how much time the other counsel wasted trying to locate the non-existent cases. I do know, however, that this issue consumed courtroom time by increasing the length of the appearance on March 13, 2026. It also consumed my time before and after that appearance.”

Borsic left the incident to the Law Society of Alberta for determination.

CTV News reached out to Olugbemi who said she has no comment on the matter ahead of the hearing.