
This article was authored by Reuben Fleming. It is provided for informational purposes and does not constitute legal advice. For legal advice please contact Reuben at 416-864-4871 or the firm at 416-224-1444. Note also: this article uses AI generated images created pursuant to OpenAI’s terms of service.
Generative artificial intelligence, or ‘gen-AI’ is taking up a lot of space in the public consciousness right now. What has stunned many is not only what this technology can do, but also how quickly it has progressed. Only 5 years ago, the notion of a computer program that could write short stories with little to no creative guidance was science-fiction to most, yet in 2025, gen-AI is creating realistic videos that appear completely real on first viewing.
While gen-AI is highly capable, some legal professionals have been a little hasty in their dependence on it, leading to significant issues.
At its current stage of development, there are a number of problems with relying on gen-AI for legal work, but perhaps the most problematic is its tendency towards ‘hallucinations’. A feature of gen-AI is that it sometimes ‘hallucinates’ when answering queries, meaning that when it does not know the answer to a question posed by the user, it may respond by presenting fictious information as fact.

This issue is only compounded by the fact that this false information often appears compelling and authoritative at first glance. In the legal context, if a gen-AI program cannot find case law or statutes that support a position, it may dream up sources with fabricated quotes and citations that appear legitimate, and only on further scrutinization does it become apparent that the cited authorities do not exist.
For instance, in Zhang v. Chen, 2024 BCSC 285, the responding party sought special costs against counsel for the moving party, following the discovery of multiple fictious citations authored by a gen-AI program which were included in the moving party’s application. In another case, Geismayr v. The Owners, Strata Plan KAS 1970, 2025 BCCRT 217, self-represented litigants cited 10 cases that they obtained using gen-AI, 9 of which were determined to be non-existent.
Gen-AI hallucinations have become a serious problem for lawyers and self-represented parties attempting to get legal assistance and advice from gen-AI programs. The presentation of non-existent cases and statutes to the court could spell disaster for a party’s position when the authorities upon which they have built their case turn to sand. Not to mention that such misrepresentations to the court could feasibly result in disciplinary action against the lawyers putting them forward.
The takeaway from all of this is not that gen-AI is bad or even unhelpful; it is an incredible tool that has a lot to offer legal professionals. However, at its current stage of development, commercially available gen-AI cannot do the job of a lawyer, a paralegal or a law clerk. Besides its propensity to hallucinate, gen-AI is nowhere close to being able to take the wheel in the provision of legal services—the law is an inherently human profession (send your letters) that requires a nuanced application of legal principles to every new situation, taking into account client needs, priorities, and concerns. While there are plenty of ways that gen-AI can contribute to the law, at present, those contributions require direction and verification by a qualified legal professional.

To access the Law Society of Ontario’s practice notes and checklists regarding the use of gen-AI, please visit the Law Society of Ontario’s Technology Resource Centre page at: https://lso.ca/lawyers/technology-resource-centre/practice-resources-and-supports/using-technology.