Quebec’s higher and lower courts have issued joint guidelines that sharply restrict how judges may use generative artificial intelligence, underlining that adjudication must remain a human task.
Human judgment remains central, courts say
The document, adopted by the Quebec Court of Appeal, Superior Court, Court of Quebec and municipal courts, permits generative AI for certain support tasks but bars it from performing core judicial functions. Those include assessing evidence, applying legal reasoning and conducting deliberation — responsibilities the courts say must stay with judges.
“Generative AI, however sophisticated, possesses neither judgment nor conscience and is not subject to any form of accountability.”
The 10-page set of guidelines states that current generative-AI tools cannot grasp the human, social and legal contexts that shape individual cases. The courts also point to the opacity of many AI algorithms and note that there is no approved institutional generative-AI system for judicial use at this time.
Limited, tightly supervised uses allowed
While the guidance draws a firm line around decision-making, it recognises potential uses of generative AI for non-substantive tasks. Judges may, under strict human control, employ AI to:
- revise, rephrase or translate text they themselves have produced;
- perform certain administrative functions;
- locate information within specified, reliable sources.
Any use must remain under the judge’s control. The courts caution that AI outputs cannot be treated as legal authority or relied on for evidentiary assessment.
Context: concerns prompted the guidance
The move follows reporting earlier this year that suggested a judge’s written decision contained references to non-existent case law, raising the possibility that generative AI had been used improperly. That episode heightened urgency within the judiciary to set rules limiting AI’s role in producing judicial work.
In their statement, the courts emphasise that judging extends beyond technical analysis. It is described as an “intrinsically human function” involving deliberation, ethical considerations, respect for dignity and the exercise of moral responsibility — elements they say AI cannot replicate.
What the guidelines mean for litigants and the public
For the public and those engaged in litigation, the rules are intended to preserve transparency and accountability in court decisions. By forbidding AI from being the source of legal reasoning or from assessing evidence, the judiciary aims to prevent errors that could arise from unverified or fabricated AI outputs.
At the same time, limited uses — such as language editing or targeted searches — may help judges and court staff manage workloads, provided safeguards remain in place to ensure human oversight.
| Adopting bodies | Role |
|---|---|
| Quebec Court of Appeal | Highest provincial appellate court |
| Superior Court | Trial-level court for serious civil and criminal matters |
| Court of Quebec | Handles provincial-level civil and criminal cases |
| Municipal courts | Local courts dealing with municipal and bylaw matters |
The document also notes that existing commercial generative-AI services rely on algorithms whose inner workings are largely opaque, which complicates any attempt to verify or attribute outputs. That opacity is another reason the courts restrict AI’s role.
Broader implications and next steps
Quebec’s judiciary joins a growing number of institutions wrestling with how to integrate AI safely. The approach is cautious: permit administrative efficiencies but block delegation of judicial functions that require moral judgment, context-sensitive interpretation and formal accountability.
The guidelines do not endorse any specific institutional tool; rather, they set principles intended to guide judges’ day-to-day decisions about whether and how to rely on generative AI. The courts will likely revisit these principles as technology and oversight mechanisms evolve.
For now, the central message is clear: AI can assist, but judges must remain the decision-makers.
