5 Key Takeaways
- AI-generated legal citations can be entirely fabricated, despite appearing plausible.
- Lawyers have an absolute ethical duty to verify every case and authority they submit to a court.
- Legal regulators are imposing increasingly severe penalties, including suspension, for misuse of AI in court filings.
- Reliance on AI without thorough human review undermines court integrity and can devastate a client's case.
- The legal profession must adapt through updated education, guidelines, and a culture of transparency around AI use.
When ChatGPT Ghostwrites a Court Case: A Canadian Lawyer's Suspension Sends a Warning to the Legal World
The first-ever suspension of a Canadian lawyer for submitting AI-fabricated case law marks a turning point in how the justice system confronts generative artificial intelligence — and raises urgent questions about competence, trust, and the limits of automation.
A Canadian lawyer has been suspended from practice for six months after being caught presenting bogus legal cases dreamed up by an artificial intelligence tool. The case marks a dramatic milestone in the uneasy marriage between the legal profession and generative AI. Mary Hyun-Sook Lee, who practises under the name Jisuh Lee, was sanctioned by the Law Society of Ontario's discipline tribunal for professional misconduct that a judge said deliberately misled the court. The tribunal ordered her to pay C$10,000 in costs and complete remedial training before she can return to work.
The episode began in a routine family and estate law hearing inside an Ontario Superior Court. Lee submitted a written legal argument called a factum — a document that lays out the facts of a case and the previous court decisions a lawyer believes should guide the judge's ruling. But when Justice Fred Myers began examining the cases Lee had cited to support her client's position, something didn't add up. Several of the supposed legal precedents seemed to have evaporated into thin air. Others existed, but said the opposite of what Lee claimed.
Justice Myers asked Lee to produce copies of the authorities she was relying on. She couldn't. So the judge did what any sceptical researcher would do: he went looking for them himself. He scoured legal databases. Some of the decisions could not be found anywhere. In one glaring example, a case that Lee presented as a reason to remove an estate trustee turned out to be a precedent that reached exactly the opposite conclusion — it supported keeping the trustee in place. The disconnect was so stark that the judge directly questioned whether artificial intelligence had been used to prepare the material.
Lee's response in court did little to reassure. She said her office did not usually rely on such tools, but she would need to check with her clerk. That answer likely raised more eyebrows than it settled. By then, the damage was done. The written reasons for the judge's decision would later lay bare the heart of the problem: a lawyer's core duty is to verify what they place before a court, and Lee had failed in that duty.
This was not an isolated slip. The tribunal that disciplines lawyers in Ontario — an independent body that governs the profession — found that Lee had submitted a factum "containing non-existent or irrelevant case law generated by an artificial intelligence tool." In plain language, she had let an AI do her homework and then handed it in as if she had checked it herself. The tribunal concluded that her conduct fell below the standard expected of a lawyer and that she misled the court.
A Marked Escalation in How Regulators Confront AI Misuse
The six-month suspension is a marked escalation in how legal regulators in Canada are confronting the misuse of artificial intelligence. Toronto lawyer Tom Macintosh Zheng, who tracks how AI is infiltrating legal practice, pointed out that this is the first time a Canadian law society has suspended a lawyer for misusing AI in court submissions. Previously, the go-to penalty for similar infractions was a reprimand — a formal scolding that stays on a lawyer's record but doesn't bar them from working. Zheng put the milestone in stark terms:
"There is increasing scrutiny on how lawyers use these tools, particularly where they affect the integrity of court proceedings."
The scrutiny Zheng describes didn't emerge from nowhere. The popularization of large language models like ChatGPT, beginning in late 2022, caught the legal world off guard. Almost immediately, lawyers began experimenting with the technology to draft motions, summarize documents, and, apparently, to generate lists of cases that looked authoritative but were pure inventions. The most infamous incident happened in a New York federal court in 2023, when an attorney used ChatGPT to write a brief and the chatbot helpfully supplied several fake rulings, complete with names, citations, and plausible-sounding analyses. The judge in that case fined the lawyers involved and the story ricocheted around the globe, becoming a cautionary tale.
Yet here, years later, a similar scenario played out in Canada. The details of Lee's case suggest that the lesson hasn't fully sunk in across the profession. The AI tool in question was not identified by name, but the characteristics — fabricated cases, citations that mix real and fake elements, confident prose that collapses under scrutiny — align with what generative AI chatbots produce when asked to find legal authority. These tools are designed to predict the next plausible word in a sequence. They are not databases of actual court rulings. They can string together convincing-sounding case names, judges, years, and legal principles that never existed. To a busy lawyer racing against a deadline, the output can look complete. But to a judge or an opposing counsel who tries to verify it, the illusion shatters.
Justice Myers's written comments from the original hearing cut to the core of a lawyer's ethical obligation. He wrote that it should be "self-evident" that counsel must read and confirm the cases they cite and must not rely on authorities that do not exist or that support the opposite of their position. This principle is not new. It predates the internet. Long before AI, lawyers were expected to shepardize their cases — a meticulous process of checking that a decision hasn't been overturned or diminished by a higher court. Technology has made that verification faster, but it has also opened a door to a new kind of shortcut: simply asking a chatbot to do the thinking for you.
The Client Caught in the Crossfire
The consequences for Lee's client are less visible but worth considering. When a lawyer cites cases that don't exist or that undermine their own argument, the client's entire legal strategy can implode. The court may lose trust in everything else that lawyer has said. In the adversarial arena of family and estate law, where emotions and stakes run high, that loss of credibility can be devastating. It could mean the difference between retaining an inheritance and losing it, or between a fair settlement and a protracted, costlier battle.
The Law Society tribunal, in handing down its penalty, didn't just punish Lee for a moment of carelessness. It required her to complete specific remedial measures during her suspension, signaling that the regulator views this not merely as an ethical lapse but as a competence issue that needs retraining. The C$10,000 in costs she must pay is more than a token sum; it is meant to drive home the financial risk of cutting corners with AI. For a sole practitioner or a small firm lawyer, that amount stings.
And Lee's legal troubles are not over. Separate from the suspension, she is still facing civil contempt proceedings. In those proceedings, the court has appointed an amicus curiae — an independent lawyer whose role is to assist the court when a party cannot or does not adequately represent themselves. The term comes from Latin meaning "friend of the court." The appointment was made, according to the court, because of procedural uncertainty, including the fact that Lee has not retained legal counsel in the contempt matter. That adds a layer of personal and professional peril. A finding of contempt could carry its own sanctions, ranging from fines to, in extreme circumstances, imprisonment.
A Signal Flare for the Profession
The Lee case is not an outlier; it is a signal flare for a profession still grappling with how to integrate powerful AI without eroding the trust that the justice system depends on. Law societies across Canada and abroad have been scrambling to issue guidelines. Many now explicitly tell lawyers that they must not submit AI-generated material without thorough human review, and that they are ultimately responsible for anything filed with the court, regardless of its origin. Some judges have started requiring lawyers to certify whether AI was used in drafting submissions and, if so, to confirm it was verified. The message is clear: AI can be a useful assistant, but the licence to practice law isn't transferable to a machine.
For the general public, the spectacle of a lawyer being suspended for citing fake cases might seem like an oddly specific sort of folly. But it touches on a broader vulnerability. As generative AI seeps into medicine, accounting, engineering, and journalism, the question repeats: who is checking the output? In law, the stakes are the fairness of a trial, the custody of children, the transfer of property, a person's freedom. The Lee suspension is a stark reminder that the humans in the loop have a job to do.
Tom Macintosh Zheng's observation that scrutiny is increasing is borne out by the timeline. Just a few years ago, a lawyer might have quietly withdrawn a dubious argument and faced little more than embarrassment. Now, a judge's pointed questioning can trigger a formal discipline case and a public ruling that makes headlines. The Ontario tribunal's written decision will now be circulated among legal ethics instructors, cited in continuing education courses, and added to the growing body of case law about what not to do with AI.
Practical Lessons for Lawyers
- Never assume AI output is factually accurate. Treat it as a starting point, not a finished product.
- Always retrieve and read the original cases yourself. If you cannot find a case the AI cited, assume it does not exist until proven otherwise.
- Maintain clear communication with your support staff. If a clerk runs a query through an AI platform, the supervising lawyer still owns the result. Blaming a clerk does not deflect the ultimate duty.
Is Legal Education Keeping Pace?
The episode also raises the question of whether legal education is keeping pace. Law schools have rushed to introduce modules on AI and legal ethics, but a generation of mid-career lawyers may have missed that training entirely. A lawyer who graduated ten years ago learned to research using tools like Quicklaw and Westlaw, which are curated human-run databases. They may not intuitively appreciate that a language model is a different beast — a statistical parlor trick of extraordinary eloquence and zero allegiance to reality. Regulators may soon need to mandate tech competency checks as part of annual continuing professional development.
Looking ahead, the Lee suspension will almost certainly not be the last of its kind. Court systems are still processing the aftermath of the initial AI boom, and more instances of hallucinated citations are bound to surface. The key factor that will determine how the profession adapts is transparency. If lawyers are honest about when and how they use AI, and if they invite a culture of open verification instead of blind reliance, the technology could enhance access to justice — helping clients who cannot afford sprawling legal research and document review. But if the technology is used in the shadows, as a time-saving black box, the outcome will be more ruined hearings, more discipline cases, and a creeping corrosion of public confidence in the courts.
As the amicus curiae continues to assist the court in Lee's lingering contempt matter, the full personal and career costs of this episode are still being tallied. A lawyer who once stood in court as an advocate now sits at the centre of a disciplinary file that will be studied for years. The image of a judge typing case names into a search engine and drawing blanks is one that should worry every professional who is tempted to outsource judgment.
The lesson from this six-month suspension is etched in unmistakable language: lawyers must know what they put before a court is real, because the opposite is a breach of the very oath that gives them the privilege to practice. The humans in the loop have a job to do — and no chatbot can stand before a judge and vouch for the truth.
No comments:
Post a Comment