Quebec’s courts have issued new guidelines restricting judges’ use of generative artificial intelligence, stressing that judging must remain a fundamentally human responsibility.
asserted
judging → issue → intelligence
The guidelines, jointly adopted by the Quebec Court of Appeal, Superior Court, Court of Quebec and the province’s municipal courts, say generative AI can be used for certain supporting tasks but cannot replace judicial reasoning, the assessment of evidence or deliberation.
asserted
AI → adopt → evidence
Judging, they say, must remain the exclusive responsibility of judges.
asserted
they → say → judges
“Generative AI, however sophisticated, possesses neither judgment nor conscience and is not subject to any form of accountability,” states the 10-page document.
asserted
document → possess → accountability
“It cannot understand the human, social and legal context specific to each case.
asserted
It → understand → case
The move comes months after questions were raised about whether a Quebec judge may have used generative AI in preparing a decision.
uncertain
judge → come → decision
La Presse reported in March that a decision contained references to case law that did not exist, raising the possibility that generative AI had been used.
asserted
AI → report → possibility
The episode helped create a sense of urgency within the judiciary as it worked to establish rules for the technology.
asserted
it → create → technology
The courts say there is currently no approved, secure institutional generative AI tool available to judges.
asserted
courts → say → judges
They also warn that existing tools rely on algorithms whose workings remain largely opaque.
asserted
workings → warn → algorithms
“Judging cannot be reduced to a technical exercise,” the guidelines say.
asserted
guidelines → reduce → exercise
“It is an intrinsically human function, based on deliberation, consideration of context, respect for dignity, the requirement of fairness and the exercise of moral responsibility.”
asserted
It → base → responsibility
The courts acknowledge that generative AI can be useful, but say it cannot be used as a source of legal authority or reasoning, to assess evidence or to make decisions.
asserted
it → acknowledge → decisions
Judges may use it for limited tasks such as correcting, revising, rephrasing or translating text they have already written, certain administrative work, or locating information within specified sources.
uncertain
they → use → sources
Any such use must remain under human control, and AI-generated material must be independently verified.
asserted
material → remain → control
The guidelines make clear that judges remain fully responsible for their legal reasoning, the contents of their decisions, and the accuracy of the facts, references, quotations and legal authorities they cite.
asserted
they → make → facts
“No error can be attributed to the tool used,” the document says.
asserted
document → attribute → tool
“The use of a generative AI tool does not diminish the judge’s obligation to verify or their full responsibility for the final result.”
asserted
use → diminish → result
The guidelines say AI cannot be used for substantive legal reasoning, analyzing or characterizing facts, assessing evidence or credibility, determining the outcome of a dispute, or drafting those parts of a decision.
asserted
AI → say → decision
“If the use of AI has the effect or purpose of determining the legal outcome, assessing credibility or interpreting disputed facts, interpreting the relevance of facts, or formulating the reasons for a decision, it must not be used,” the guidelines say.
asserted
guidelines → have → decision
The courts also warn judges about confidentiality risks.
asserted
courts → warn → risks
Judges should not provide AI tools with draft judgments, notes about ongoing cases or information protected by law or a court order.
asserted
Judges → provide → law
Until an authorized institutional system is available, judges using an AI tool for permitted purposes must also ensure its settings prevent information they enter from being used to train or improve the system.
asserted
they → use → system
When in doubt, the guidelines recommend judges refrain from using generative AI.
asserted
judges → recommend → AI
The guidelines are intended to protect principles including judicial independence, impartiality, fairness, human rights and access to justice, while ensuring confidentiality, security and data protection.
asserted
guidelines → intend → confidentiality
The Aporia
