Canada’s national AI strategy leaves out the justice system
Somewhere in Canada this year, a self-represented litigant filed court decisions to support their case. The names, the courts, the years all read like real law. They were not. An AI chatbot had invented them, and the litigant handed them to a judge as binding authority.
This has now happened at least 172 times across 53 Canadian courts and tribunals, according to Courtready’s tracking of published decisions. In 81 per cent of those cases, the person who filed the fabricated material was representing themselves.
Two facts explain why this matters. AI chatbots invent court cases: asked for authority, they will produce—with complete confidence—decisions that do not exist. And Canadian courts run on precedent. The system assumes the cases placed before a judge are real. A fabricated decision is, therefore, not just wasted court time. If it slips through, it can be cited in argument and—in the worst case—folded into a judge’s reasons where it becomes a citable authority in its own right. That has already happened in a Canadian decision. Once a fake case is cited in a real one, it becomes part of the common law. The reliability of our legal system depends on this not happening.
This is the problem Canada’s national AI strategy, AI for All, does not address. In fact, it does not appear to address the justice system at all. The strategy is sweeping: privacy, elections, health care, education, the labour market, the public service. When it names where AI already makes consequential decisions about people’s lives, it lists hiring, lending, health care, and public services. The courts, where the stakes are a person’s liberty, money, housing, or children, do not appear. It mentions the rule of law once, as a national strength, and legal analysis once, as a task AI can perform. Neither treats the justice system as a place where AI is used, creates risk, or needs governing. For a harm already documented 172 times, that silence is hard to defend.
The people most exposed are the ones the strategy says it most wants to protect. Self-represented litigants are a large share of those in Canadian courts, and Canadians who cannot afford a lawyer increasingly turn to free AI tools to build their cases. They are the least equipped to recognize a fabricated citation, and the 81 per cent figure is the predictable result. It is the same literacy gap the strategy itself flags, ranking Canada 44 of 47 countries. The difference is that in a courtroom, being AI-illiterate carries formal legal consequences.
Those consequences are real, from cautions to costs orders, and the litigants who bear them have no insurer or law firm to absorb the loss. How courts sanction this conduct—and how consistently—is for the judiciary and tribunals to resolve. A federal AI strategy cannot dictate that. But whether Canadians are equipped to avoid the problem in the first place is squarely within the strategy’s reach, and that is where its silence matters.
The fix does not require reopening the strategy. It requires extending what the strategy already does. AI for All commits to a national AI literacy effort delivered through schools, workplaces, and the public libraries it calls trusted community hubs.
Self-represented litigants need that same effort aimed at them: plain-language guidance, at the courthouse, the legal clinic, and online, built around the one rule that matters most in legal research, that every case must be checked against the official source before it is filed. Free public databases and verification tools already make that check possible. What is missing is the effort to put them in front of the people who need them, when they need them.
Canada’s AI strategy sets out to map where AI touches people’s lives and prepare Canadians to use it safely. It has reached nearly every such place except the courtroom, already the site of a documented, repeating, and largely preventable harm. Bringing self-represented litigants into that literacy agenda is a small step. Leaving them out means the Canadians most exposed to AI’s failures in court, and least able to afford a lawyer to catch them, are left to face the problem alone.
Tom Macintosh Zheng is a Toronto-based lawyer and co-founder of Courtready where he conducts research and builds tools, including with AI, to advance access to justice in Canada. He is this year’s winner of the Ontario Bar Association’s OBA Foundation Award, which recognizes exceptional contributions to the improvement of the justice system through public legal education, innovative research, or other means.
The Hill Times