Not long ago, a sudden shift in a client’s writing style was easy to explain. If someone who usually sent short, to-the-point emails suddenly produced a long, polished message that read like a legal brief, chances were a spouse, sibling, or friend had helped. Today, that assumption no longer holds.
Now, as Toronto family lawyer Ron Shulman has discovered, the explanation is far more likely to be artificial intelligence.
The rise of generative AI has quietly but decisively entered the legal system, changing how clients communicate with lawyers and, increasingly, how people attempt to represent themselves in court. While the technology promises efficiency and access to information, its growing use in legal matters is exposing serious risks for clients, lawyers, and the justice system itself.
On the surface, the appeal is obvious. AI tools can summarize dense information, organize thoughts, and even draft documents in seconds. For people overwhelmed by legal processes or deterred by high legal fees, AI feels like a shortcut to empowerment. But that sense of confidence can be dangerously misplaced.
Lawyers across Canada are reporting a surge in AI-generated materials landing on their desks. Clients arrive with lengthy submissions, research memos, or even completed applications created by AI and expect them to be filed as-is. In some cases, clients rely on AI not just for drafting, but for deciding legal strategy treating it as an all-knowing authority rather than a fallible tool.
That’s where the trouble begins.
AI systems are known to “hallucinate,” confidently inventing cases, laws, or legal pathways that simply do not exist. Courts have already seen the consequences. Lawyers and self-represented litigants alike have been sanctioned for submitting fake authorities generated by AI. In one particularly serious case, a Toronto lawyer is now facing criminal contempt proceedings after filing AI-invented cases and denying their origin when questioned by a judge.
For self-represented litigants, the financial fallout can be severe. Courts in Quebec and Alberta have imposed monetary penalties on individuals whose AI-assisted submissions cited non-existent or irrelevant legal authorities. Far from saving money, these missteps have increased costs and delayed proceedings sometimes for everyone involved.
Even when penalties aren’t imposed, AI can strain the lawyer-client relationship. Immigration lawyer Ksenia Tchern McCallum describes clients who run her work through AI to “fact-check” it, unknowingly risking their own privacy by uploading sensitive documents. Others challenge her professional advice because an AI tool suggested something different. The result is friction, mistrust, and a breakdown in effective representation.
These dynamic highlights a core misunderstanding: AI can scan the internet and reproduce common patterns, but it cannot replace professional judgment, contextual understanding, or ethical responsibility. It doesn’t know which arguments resonate with a particular judge, which details are strategically risky, or which procedural nuances could derail a case.
Ironically, the more confident AI sounds, the more misleading it can be. As Shulman recounts, clients sometimes send pages of AI-generated legal analysis that are entirely irrelevant to their situation. Lawyers still have to read and correct everything time that clients ultimately pay for.
To be clear, this is not an argument against AI itself. Used carefully, it can be a powerful aid. It can help people understand general processes, organize timelines, or prepare questions for their lawyer. Some legal professionals already use advanced, paid AI tools behind the scenes to improve efficiency, always with human review and strict privacy safeguards.
The problem lies in uncritical reliance.
As Jennifer Leitch of the National Self-Represented Litigants Project puts it, this moment is a kind of “harm reduction” phase. People are going to use AI whether the legal system likes it or not. The challenge is teaching responsible use: verifying every citation, following court rules, disclosing AI assistance when required, and understanding the limits of what these tools can do.
AI has real potential to improve access to justice, especially for those who feel shut out by complexity and cost. But right now, the legal landscape is something of a Wild West. Open-access tools are uneven in quality, and the confidence they project often masks serious inaccuracies.
In the end, AI is just that a tool. It doesn’t carry ethical duties. It doesn’t bear the consequences of being wrong. And it doesn’t understand the human realities behind legal disputes.
Courts run on judgment, accountability, and trust. Until AI can offer those and it can’t it should assist the legal process, not attempt to replace the people responsible for navigating it.
