Updated document aims to help federal judiciary navigate AI tools, deepfakes
The Canadian Judicial Council, the body that oversees Canada’s federal judiciary, released two sets of guidelines on Tuesday to help judges navigate artificial intelligence and evaluate digital media evidence, including potential deepfakes.
The CJC released the first edition of its AI guidelines in 2024, stating that its purpose was to raise awareness of AI risks while encouraging the judiciary to explore “safe, effective, and appropriate uses” of AI tools. Multiple courts across the country say that guidance has shaped their judges’ approaches to generative AI.
In comparison, the 2026 guidelines state that their objective is to encourage judges to leverage AI to support the administration of justice while ensuring that their use of AI aligns with judicial principles.
The second set of guidelines released Tuesday, which focuses on evaluating digital media evidence, aims to help judges in cases where evidence is alleged or suspected to have been created or altered by AI tools. The guidelines include checklists that judges can use to evaluate so-called acknowledged, unacknowledged, and indirect AI evidence.
Tom Macintosh Zheng – a former commercial litigator whose organization, Courtready.ca, has tracked court cases involving AI-hallucinated case citations – says the 2026 AI guidelines lay out, “in a very useful way, what tasks can or should judges use AI for and what tasks they should not be using AI for.” The appendix of the document provides specific examples of appropriate, high-risk, and inappropriate uses of AI.
“In my conversation with lawyers, with adjudicators, this is the same question they’re asking me,” Macintosh Zheng says.
He argues that the CJC’s shift from raising awareness of the risks of AI in 2024 to encouraging AI adoption with explicit guardrails in 2026 is significant because AI tools could help courts work more efficiently. “Why is it that the benefits of artificial intelligence should only be limited to Bay Street?” Macintosh Zheng asks.
“I think judges and court administrators, having now read this document, will do better at identifying places where AI can truly be an advantage for judicial practice.”
Kathryn Manning, a litigator and founding partner of DMG Advocates, notes that the rapid evolution of AI tools has introduced new challenges around evaluating the authenticity of evidence, making the new guidelines for evaluating digital media evidence particularly helpful.
“I think as a lawyer, we want to know that the judges we’re appearing in front of have the tools to deal with this really quickly evolving area,” Manning says. “Nobody can be an expert in everything, and dealing with technology and the changes in it is something that wasn’t part of their day job a few years ago.”
Below are some of the biggest takeaways from the CJC’s new guidance.
AI guidelines say courts should be transparent about AI use
According to the AI guidelines, the use of AI tools by an individual judge will not typically require disclosure any more than the use of other tools and support staff, such as law clerks, legal researchers, translation software, or spelling and grammar checkers. In cases where AI is used only for administrative, organizational, or editorial assistance, responsibility for the work remains solely with the judge.
However, as institutions, courts must be transparent about which AI tools they use and how they use them. The guidelines state that courts “should be prepared to explain how AI systems have been evaluated, approved, governed and secured; the safeguards that apply to their use; the types of functions they perform; and the oversight mechanisms established to ensure that their use remains consistent with judicial independence, procedural fairness and the proper administration of justice.”
In Manning’s view, this approach strikes an appropriate balance between transparency and judicial independence.
“The public needs to feel confident in how courts are assessing and implementing these types of technology,” she says. But at the same time, we can’t be asking judges to tell us how they were writing their decisions.”
Macintosh Zheng notes that a growing number of tribunals across Canada are requiring litigants to disclose AI use. “I think if the parties who appear before the courts are required to disclose such use, then it also makes sense that courts are also disclosing it,” he says.
Evidence guidelines flag risk that real evidence will be accused of being fake
In its guidance on evaluating digital media evidence, the CJC notes that discussions of AI-fabricated evidence typically focus on the fear that such evidence will be admitted into legal proceedings.
However, the CJC flagged an “equally significant” risk: “the possibility that authentic evidence will be successfully challenged as fabricated, and thus excluded from evidence.” This phenomenon has been referred to as both the “deepfake defence” and the “liar’s dividend.”
“In either case,” the CJC added, “the result may lead to a miscarriage of justice.”
Manning says the latter phenomenon is reminiscent of situations in which real news is accused of being “fake news.” She adds, “I think that is an important thing for judges and lawyers to be aware of.”
The CJC noted that Canadian courts have already begun grappling with this issue. In a case called R. v. Medow, the Ontario Court of Justice rejected last year unsupported allegations that police bodycam footage had been manipulated.
“The decision confirms the principle that courts should neither ignore the reality of synthetic media nor assume that every allegation of manipulation warrants extensive inquiry,” the CJC said.
The checklist for evaluating unacknowledged AI evidence provides step-by-step guidance on this issue.
Evidence guidelines raise access to justice concerns for self-represented parties
The guidance on evaluating digital media evidence also raises concerns that allegations of fabricated evidence will undermine access to justice for self-represented and low-income parties.
These litigants are less likely to have the resources to retain forensic experts to analyze digital evidence or evaluate synthetic media, the CJC said; judges should therefore “remain attentive to fairness and access to justice considerations in such circumstances.
“While courts must maintain impartiality, they may consider appropriate procedural measures to ensure that allegations of manipulation receive genuine scrutiny without allowing unsubstantiated claims to consume disproportionate resources or undermine trust,” the CJC added.
Macintosh Zheng agrees that access-to-justice issues could be on the horizon. He provides an example of how the court’s concerns could play out.
“Let’s say you’re a landlord, and all of a sudden, you get an image [from your tenant] saying, ‘Oh, the fridge is broken. Can you please send me $500 to fix it?’”
A landlord with the means to hire a forensic expert could have that expert opine that that image is fake, Macintosh Zheng says. But the tenant may not have the money to hire a comparable expert to argue that the image is authentic.
He notes that the CJC’s guidance does not offer a substantial solution to this issue.
“I think the mechanism is something that the legal community should work on, because there’s going to be that gap between those who can afford experts… and those who won't have that capability,” he says.
