Ontario Civil Rules Committee: Proposes New AI Evidence Rules

• Ontario is implementing new civil procedure rules to manage AI-generated evidence in its courts, developed by the Ontario Civil Rules Committee AI, chaired by Justice Peter Lauwers.
• The rules differentiate between acknowledged and unacknowledged AI content, with the latter, including deepfakes, facing a presumption of inadmissibility.
• Courts may require disclosure of AI systems, training data, inputs, and prompts for acknowledged AI evidence, with bias as a key concern for AI-generated evidence admissibility Ontario.
• Prompt disclosure is discretionary for judges, balancing transparency needs against solicitor-client privilege, impacting whether evidence is admitted.
• The regulations aim to address current issues like AI hallucinated citations Ontario and passive reliance on AI observed in legal proceedings.

This robust stance underscores the judiciary's commitment to preventing fabricated evidence from undermining the adjudicative process, while still allowing for the strategic use of AI when proper transparency is maintained.

Ontario is poised to introduce new civil procedure rules specifically designed to govern the use of AI-generated evidence within its court system. This initiative responds to growing concerns about the integrity of legal submissions, including issues like fabricated case citations, AI-modified quotations, and an over-reliance on artificial intelligence tools, which are already presenting challenges in Canadian courtrooms. The proposed framework, developed by the Ontario Civil Rules Committee (CRC) under the leadership of Justice Peter Lauwers of the Court of Appeal for Ontario, adopts a strategy of judicial oversight rather than outright prohibition of AI tools.

The core philosophy behind these new rules for Ontario civil procedure AI evidence is to ensure transparency and allow for rigorous scrutiny of AI-produced material, while granting judges discretion over specific disclosures, such as prompts. Justice Lauwers, who also chairs the CRC's subcommittee on artificial intelligence, oversaw a consultation process with legal professionals in 2025. Following this feedback, a significantly revised set of rules was prepared and has since been submitted to Attorney General Doug Downey. The legal community anticipates the release of the finalized rule text in November, marking a significant step towards formalizing how AI interacts with the judicial process.

The forthcoming regulations establish a clear distinction between two categories of AI-generated content. The first category addresses acknowledged AI-generated evidence, which encompasses any material explicitly disclosed as having been produced, either entirely or in part, by an artificial intelligence system. For such submissions, a court may compel the party presenting the evidence to identify the specific AI system employed, detail its training data, outline the inputs and prompts used, and provide supporting documentation to attest to its validity and reliability. A primary focus for the courts will be to identify and mitigate potential biases, including sampling bias and discriminatory bias, particularly those prohibited under the Ontario Human Rights Code.

The second category targets unacknowledged AI-generated evidence, a critical area that includes sophisticated fabrications like deepfakes. Given the persuasive nature and inherent difficulty in discrediting artificially created video or audio content, the proposed rule establishes a presumption of inadmissibility for any material discovered to be undisclosed AI-generated content. This robust stance underscores the judiciary's commitment to preventing fabricated evidence from undermining the adjudicative process, while still allowing for the strategic use of AI when proper transparency is maintained. This approach, as articulated by Justice Peter Lauwers, prioritizes judicial control over an outright ban, ensuring that courts can demand sufficient transparency to evaluate AI-generated material effectively.

A particularly complex aspect of the new framework concerns the disclosure of prompts used to generate AI content. The way a question or instruction is phrased can significantly influence and potentially bias the output of an AI system. For instance, asking an AI to assess an individual's "lack of productivity" versus their "productivity compared to peers" could yield vastly different results. This inherent sensitivity makes prompt disclosure a critical consideration for AI-generated evidence admissibility in Ontario courts.

However, prompts can also contain information that reveals a party's litigation strategy, potentially falling under solicitor-client privilege. Recognizing this, the proposed rules do not mandate prompt disclosure as an automatic right. Instead, the decision to compel such disclosure rests entirely with the presiding judge's discretion. Justice Lauwers has indicated that while lawyers may not be forced to reveal their prompts, a refusal to do so could lead to the evidence being deemed unreliable or invalid, and consequently, inadmissible, if the prompt is perceived to have unduly distorted the AI's output.

The need for these new regulations is underscored by existing issues already surfacing in Ontario courtrooms. Justice Lauwers has highlighted that common problems at the appellate level include incorrectly generated legal references, often referred to as AI hallucinated citations Ontario, and AI-edited quotations. In response, rules have already been modified to require legal professionals to certify the accuracy of citations and other submitted materials. The broader concern, however, extends to the passive reliance on AI tools.

A 2025 MIT study, cited by Justice Lauwers, demonstrated that individuals who initiate tasks with AI assistance, rather than their own reasoning, often struggle to articulate the basis of the AI's answer just minutes later. Justice Lauwers has observed this pattern among counsel when pressed by the bench, indicating a potential erosion of independent legal thought. Kathryn Manning, a litigation partner at DMG Advocates in Toronto and a member of the CRC's AI subcommittee, confirms that this same dynamic is evident at the trial level, emphasizing the widespread implications of these evolving AI rules for lawyers practicing in Ontario. The rules aim to mitigate these risks by ensuring that deepfakes evidence Ontario courts might encounter is handled with extreme caution, and that all AI-generated content is subject to rigorous judicial review.

Lawyers practicing in Ontario civil courts must prepare for new rules governing AI-generated evidence, requiring transparency regarding AI systems, training data, and prompts, and facing a presumption of inadmissibility for unacknowledged AI content like deepfakes. Firms should review their internal AI usage policies to ensure compliance with these forthcoming judicial scrutiny standards.

Source: Original reporting via Law Times