Regardless of opinions, feelings, or fears, AI is here and legal practitioners need to adapt our systems accordingly. The use of AI in the practice of law presents opportunities but also challenges. The same is true for the courts. This has been recognized by the Canadian Judicial Council (the “CJC”), which is the national body that oversees federally appointed judges in Canada. Last month, the CJC released a report titled Guidance for Canadian Judges on Evaluating Digital Media Evidence (the “Guidelines”). This is its second set of guidelines (the first was in 2024) for courts regarding the use of AI in the court system. The Guidelines primarily focus on how the courts can address concerns that evidence is AI generated or manipulated. It suggests that our general principles of evidence are still very much applicable: is the evidence authentic, is it reliable, and what weight should it be given.
In situations where use of AI is disclosed by the parties, such as for translation or image enhancement, the Guidelines suggest that the court’s focus will be determining the reliability of the evidence. However, the larger concern is where there are allegations that evidence is AI generated or manipulated but is presented as human-created. The admission of fake evidence could bring the administration of justice into disrepute and lead to a miscarriage of justice, but so could the exclusion of real evidence where it is alleged to be AI generated or manipulated and it is not. The issue in both cases is the authenticity of the evidence.
The courts (and this author suggests, also counsel) may need to use other available tools where authenticity is at issue. The Guidelines suggest that the courts examine the origin of the evidence, native files and metadata, chain of custody of the evidence, preservation history, expert evidence, and signs of manipulation. The Guidelines suggest that the practical process for engaging in the type of necessary examination of evidence may be one or all of the following: (i) conduct a voir dire on authenticity or admissibility; (ii) order production of original files, metadata, or related technical information; (iii) require additional witness evidence; (iv) request expert forensic analysis to resolve technical disputes; and (v) make preservation orders to preserve original files, metadata, and other primary sources. The courts will need to restrict or exclude evidence where appropriate and should provide detailed reasons regarding authenticity and reliability.
Another point raised by the Guidelines is for the courts to consider the “accountability” of the evidence. The Guidelines suggest that the accountability issue arises in situations of “indirect AI”. Indirect AI is where AI is used in the preparation or formulation of evidence, but the AI generated output itself is not being tendered as evidence. In such cases, the human witness is still accountable for the content of the evidence. Examples include where AI has been used to assist in drafting affidavits, witness statements, expert reports, or chronologies. The human witness who is adopting the evidence must understand the contents and be able to “stand behind” it. The Guidelines suggest that someone needs to take responsibility for the authenticity, accuracy and provenance of the evidence. That alone will not be enough to ensure authenticity but it may assist in acting as an initial filter if someone is answerable for the proffered evidence.
The Guidelines point out that while the use of AI is now widespread, easily accessible, and affordable (or free), the self-represented and economically vulnerable litigant will not necessarily be in the position to challenge suspected AI generated or manipulated evidence. This presents another challenge for the courts who remain the ultimate gatekeepers of the admissibility of evidence. The Guidelines end with checklists for responding to digital media evidence.
Thanks for reading!
