The province’s highest courts have moved to curb the temptation of letting machines shape legal judgments, issuing a joint directive that insists judging remains a wholly human endeavour.
New Guidelines Released
The Quebec Court of Appeal, Superior Court, Court of Quebec and the municipal courts jointly adopted a 10‑page document that sets out strict boundaries for generative artificial intelligence. While the text acknowledges that AI can assist with routine chores, it makes clear that any function that touches on legal reasoning, evidence assessment or decision‑making is off‑limits.
The guidelines quote the document directly: “Generative AI, however sophisticated, possesses neither judgment nor conscience and is not subject to any form of accountability.” It adds, “It cannot understand the human, social and legal context specific to each case.”
Background: The La Presse Incident
The push for formal rules gained momentum after a March report by La Presse highlighted a decision that contained references to non‑existent case law. The article suggested that a judge might have relied on a generative AI tool to draft the passage, sparking concern within the judiciary about the technology’s reliability.
In response, the courts stressed that no approved, secure institutional AI system is presently available to judges. They warned that existing platforms operate with opaque algorithms, making it impossible to scrutinise how outputs are generated.
Permitted Uses and Safeguards
Judges may employ AI for limited, administrative tasks such as correcting, revising, rephrasing or translating text they have already written, conducting routine clerical work, or locating information within pre‑approved sources. Any AI‑generated material must be independently verified and remain under direct human supervision.
The directive emphasises accountability: “No error can be attributed to the tool used,” the document states. “The use of a generative AI tool does not diminish the judge’s obligation to verify or their full responsibility for the final result.”
Substantive functions are expressly prohibited. The guidelines warn that AI must not be used for “determining the legal outcome, assessing credibility or interpreting disputed facts, interpreting the relevance of facts, or formulating the reasons for a decision.”
Confidentiality and Accountability
Confidentiality receives particular attention. Judges are instructed not to feed AI tools with draft judgments, notes on ongoing cases, or any information protected by law or a court order. Until an authorised institutional system is in place, users must configure their tools to ensure that entered data cannot be harvested to train or improve the model.
When uncertainty arises, the advice is simple: refrain from using generative AI altogether. The overarching aim is to safeguard judicial independence, impartiality, fairness, human rights and access to justice, while upholding confidentiality, security and data protection.
Why it Matters
By drawing a firm line between administrative assistance and judicial reasoning, Quebec’s courts reinforce the principle that justice cannot be outsourced to algorithms. The move not only protects litigants from opaque, potentially biased outputs but also preserves public trust in a system where decisions are meant to reflect human moral responsibility, not machine‑generated patterns. As other jurisdictions grapple with similar dilemmas, Quebec’s approach offers a clear template for integrating technology without compromising the core of the judicial function.