Ontario: Proposed AI Evidence Rules Address E-Discovery

• Artificial intelligence is rapidly making the creation of fake evidence easier, outpacing the development of case law.
• Ontario's proposed AI evidence rules will require proponents to prove authenticity if evidence is reasonably suspected of being AI-generated, and undisclosed AI evidence will be presumptively inadmissible.
• The Canada Evidence Act's presumption of authenticity for electronic evidence is deemed insufficient for modern deepfakes.
• The Civil Rules Review Working Group proposes shifting from relevance-based to reliance-based documentary disclosure, potentially moving the e-discovery burden earlier in litigation.
• Courts may require identification of AI systems and prompts used, with non-disclosure of prompts potentially leading to evidence inadmissibility if reliability is compromised.

While the disclosure of prompts would generally remain at a judge's discretion, Justice Lauwers emphasized a critical caveat: failure to disclose prompts could lead to the evidence being deemed inadmissible, especially if the prompts are believed to have distorted the evidence to the point of making it unreliable or invalid.

The realm of electronic discovery is undergoing a profound transformation, driven by rapid technological advancements and corresponding shifts in legal frameworks. These changes are altering both the nature of questions litigators must address and the timing within a case when these inquiries become critical. A significant factor in this evolution is the proliferation of artificial intelligence, which has simplified the creation of fabricated evidence.

This burgeoning risk is developing at a pace that significantly outstrips the evolution of established case law, as noted by Emily Hazlett, a partner at Blake, Cassels & Graydon LLP. The current legal landscape, particularly concerning evidence, struggles to accommodate the sophisticated technological capabilities now available for generating highly convincing, yet potentially inauthentic, digital content. Kelly Friedman, an e-discovery specialist and litigator at Heuristica Discovery Counsel, highlights this disparity, suggesting that existing evidence law is not adequately equipped for the current level of technological expertise that enables the creation of virtually any faked material.

A critical challenge arises from the Canada Evidence Act, which currently presumes the authenticity of electronic evidence if the system used for its storage was operating correctly. However, this standard, according to Friedman, can be easily met by sophisticated deepfakes, thereby undermining the presumption's protective intent. This gap necessitates a heightened focus on the provenance of digital materials.

Litigators are now advised to rigorously question the origin of crucial audio, video, or photographic evidence, seeking to understand its source, whether the original has been preserved, and what associated metadata exists. This proactive approach is essential given the ease with which digital content can be manipulated. Further underscoring these concerns, the Canadian Judicial Council digital media guidelines, released in October 2026, specifically address the reciprocal risk: that genuinely authentic evidence could be erroneously dismissed as fabricated. This highlights the dual challenge of discerning real from fake, and protecting real evidence from unfounded challenges, particularly in Canadian e-discovery AI authenticity contexts.

In response to these emerging challenges, Ontario's AI evidence rules for e-discovery, which include amendments to the Rules of Civil Procedure effective September 1, 2026, and further revised text anticipated in November 2026, are set to significantly impact litigation practices. Justice Peter Lauwers of the Ontario Court of Appeal, who chairs the Ontario Civil Rules Committee, outlined these rules, which aim to address the issue of AI-generated evidence from the judiciary's perspective. Under these new guidelines, if a challenger can demonstrate, on a balance of probabilities, that specific evidence could plausibly have been created by artificial intelligence, the proponent of that evidence would then be required to provide proof of its authenticity.

A key provision states that any AI-generated evidence that has not been disclosed would be considered presumptively inadmissible in court. Furthermore, if the use of AI is acknowledged, a court may compel the proponent to identify the specific AI system employed and the prompts utilized to generate the evidence. While the disclosure of prompts would generally remain at a judge's discretion, Justice Lauwers emphasized a critical caveat: failure to disclose prompts could lead to the evidence being deemed inadmissible, especially if the prompts are believed to have distorted the evidence to the point of making it unreliable or invalid.

Beyond authenticity, the timing and scope of disclosure in e-discovery are also undergoing significant reevaluation. The Civil Rules Review Working Group, in its December 2025 final report, has put forth a proposal to transition from a relevance-based model of documentary disclosure to a reliance-based approach. This new model would require parties to produce documents they intend to rely upon, as well as those they know the opposing side would require for their case.

Kathryn Manning, a litigation partner at DMG Advocates in Toronto, suggests that this shift may not necessarily alleviate the overall e-discovery burden in Canada. She anticipates that lawyers will still need to undertake comprehensive document collection from clients to determine which materials will be relied upon. Manning predicts that if this proposal is fully adopted, it will likely accelerate the e-discovery burden shift in Canada, pushing the bulk of e-discovery work to an earlier stage in the litigation process. Concurrently, generative AI is also facilitating earlier insights into cases. While e-discovery tools traditionally focused on culling vast document sets for attorney review, the emphasis is now moving towards extracting early insights and enabling a preliminary assessment of a case. Despite these technological advancements, Manning stresses the continued necessity of human oversight, including training and monitoring, for these sophisticated tools.

Lawyers and compliance officers in Canada, particularly Ontario, must prepare for significant changes in e-discovery practices due to AI-generated evidence. They should anticipate heightened scrutiny of digital evidence authenticity, understand the implications of Ontario's proposed AI evidence rules regarding disclosure and admissibility, and be aware that the e-discovery burden may shift earlier in the litigation process, requiring proactive strategies for evidence collection and assessment.

Source: Original reporting via Canadian Lawyer, Law Times, and Lexpert.