The Deepfake Defense in Canadian Civil Litigation: A Pre-Trial Playbook
R. v. Medow established that Canadian courts recognise the deepfake threat. This pre-trial playbook covers the Liar's Dividend tactic, section 31.1 of the Canada Evidence Act, Ontario Reg 384/24, and what litigators must do before trial.
A litigation tactic is now appearing in Canadian civil proceedings with increasing frequency.
It does not require fabricating evidence. It requires only raising the possibility that authentic evidence was fabricated.
Researchers Chesney and Citron named this the Liar's Dividend in 2019. As deepfake technology becomes widely known, anyone can plausibly claim that genuine video, audio, or documentary evidence is AI-generated. The existence of the technology becomes the argument. The authenticity of real evidence becomes, for the first time in legal history, a live question that opposing counsel can raise without producing any actual proof of fabrication.
Canadian courts have acknowledged the threat. Canadian procedural rules have begun to respond. And the consequences of being unprepared, whether your authentic evidence is challenged as synthetic or your opponent's fabricated evidence goes unexamined, are now severe enough that pre-trial forensic authentication is no longer optional for any file where digital media is central to the case.
This playbook covers the current state of Canadian law, the global precedents that will be cited in Canadian courts, and a pre-trial protocol for each of the three scenarios you will encounter.
R. v. Medow, 2025 ONCJ 661: Justice Brock Jones took judicial notice of the widespread proliferation of AI technology capable of producing realistic deepfake videos, describing them as highly deceptive and a serious concern to the integrity of the justice system.
R. v. Aslami, 2021 ONCA 249: Justice Nordheimer observed that certain digital evidence may require expert authentication because there are too many ways for someone to make electronic evidence appear to be something other than what it is.
The Two-Sided Problem
The deepfake problem in litigation presents two distinct failure modes. Competent pre-trial preparation must account for both.
The first failure mode is a false positive: authentic evidence is excluded, discounted, or never tendered because opposing counsel has successfully cast doubt on its integrity. In Huang v. Tesla (Cal. Super., Santa Clara, 2023), Tesla's lawyers argued that video evidence of Elon Musk making statements about Autopilot safety might be a deepfake and sought its exclusion. The court rejected the argument. The presiding judge described the tactic as deeply troubling. But the motion consumed litigation resources and required the opposing party to answer a challenge that would not have existed five years earlier.
The second failure mode is a false negative: fabricated evidence is not identified and enters the record as authentic. In Mendones v. Cushman and Wakefield (Cal. Super., Alameda County, September 9, 2025), the first reported case in which a deepfake was submitted as authentic court evidence, Judge Victoria Kolakowski identified the fabrications through visible technical tells and a metadata contradiction. The judge noted explicitly that she lacked the time, funding, and technical expertise to examine every suspicious submission in the record. Several exhibits went unexamined.
Both failure modes are live risks in Canadian civil litigation today. Pre-trial forensic authentication is the mechanism that closes both.
Chesney and Citron (2019): As deepfake technology becomes publicly known, the existence of the technology provides cover for anyone wishing to deny authentic evidence. A party does not need to prove media was fabricated. They only need to raise the possibility convincingly enough to create doubt. The technology itself becomes the argument.
Where Canadian Courts Stand
Canadian jurisprudence on deepfake evidence has moved in the last two years from theoretical concern to active judicial commentary.
R. v. Medow, 2025 ONCJ 661. Justice Brock Jones of the Ontario Court of Justice took judicial notice of the widespread proliferation of AI technology capable of producing realistic deepfake videos, describing such media as highly deceptive and noting that they present a potentially serious concern to the integrity of the justice system. Justice Jones cautioned that courts must ensure the authentication voir dire required for digital evidence is not rendered meaningless. The case involved a challenge to video evidence of an altercation. Justice Jones ultimately declined to infer that the video had been digitally altered, but his framing of the authentication standard is significant for all subsequent proceedings.
R. v. Aslami, 2021 ONCA 249. Justice Nordheimer of the Ontario Court of Appeal observed that certain digital evidence may require expert authentication because there are too many ways for an individual, who is of a mind to do so, to make electronic evidence appear to be something other than what it is. This observation, made before the current generation of consumer-grade AI tools existed, is now even more directly applicable.
Breton v. Ministry of Health and Social Services (Commission d'acces a l'information du Quebec). The Commission acknowledged that AI-generated evidence can mislead courts and stakeholders but admitted certain AI-generated documents on the basis of corroborating source material. This illustrates the variable standard Canadian tribunals are reaching.
R. v. Kapoor, 2025 ONCJ 542. An Ontario judge found that distributing a deepfake intimate image fell outside the current terms of section 162.1 of the Criminal Code as drafted. The decision illustrates both the pace at which deepfake cases are reaching Canadian courts and the gap between the current legislative framework and the technology it needs to address.
Canadian courts are treating deepfake risk as a judicially noticeable reality, but the authentication framework has not changed. The burden under section 31.1 of the Canada Evidence Act remains on the party tendering digital evidence. What has changed is the environment in which that burden is assessed.
The Global Cases Canadian Litigators Need to Know
The following decisions will be cited persuasively in Canadian civil proceedings. They are not binding authority, but they represent the most developed judicial thinking on deepfake evidence available.
Huang v. Tesla (Cal. Super., Santa Clara, 2023). The canonical Liar's Dividend case. Tesla's counsel sought to exclude a video of Elon Musk by suggesting it might be a deepfake. The court rejected the argument and described it as deeply troubling. This decision stands for the proposition that a bare deepfake allegation without forensic foundation will not succeed.
Mendones v. Cushman and Wakefield (Cal. Super., September 9, 2025). The first reported terminating sanctions case for actually-fabricated court evidence. The exhibits were identified through visible tells and a metadata-hardware mismatch. A file claimed to be recorded on an iPhone 6 running iOS 12.5.5. The AI generation feature used requires an iPhone 15 Pro and iOS 18. The court issued a terminating sanction. The ruling stands for the proposition that courts will sanction fabricated evidence submission and that metadata examination is now an expected part of evaluating suspicious submissions.
Wisconsin v. Rittenhouse. Defense counsel argued that prosecution use of pinch-to-zoom enhancement of video evidence constituted AI manipulation. The argument failed. Courts have consistently declined to treat standard digital processing as synthetic generation.
United States v. Reffitt (January 6 prosecution). Defense raised a deepfake challenge to video evidence. The court required the party raising the challenge to provide an evidentiary basis for the allegation. The challenge failed.
Courts are receptive to deepfake challenges that are forensically grounded and skeptical of those that are not. A deepfake allegation without forensic support will not succeed in a well-run court. But a deepfake allegation with forensic support, or against evidence that has not been forensically authenticated, creates real procedural risk.
The Canadian Legal Framework: Four Provisions Every Litigator Needs to Know
Before the pre-trial protocols, these are the four legal anchors that govern deepfake evidence challenges in Canadian civil proceedings.
Section 31.1 of the Canada Evidence Act — The Authenticity Burden
This provision requires the party tendering electronic evidence to produce evidence capable of supporting a finding that the electronic document is what it purports to be. The threshold is low — courts have consistently held that authentication requires only a minimal evidentiary basis sufficient to support a finding, not proof beyond a reasonable doubt. However, the threshold is not zero. A deepfake challenge that puts the authenticity question squarely in issue will require the tendering party to produce more than a bare assertion that the file is genuine. The provincial equivalents, including Ontario Evidence Act section 34.1, operate on the same framework.
R. v. Mohan and White Burgess — The Expert Evidence Framework
R. v. Mohan [1994] 2 SCR 9 established the four-part test for expert evidence admissibility in Canada: relevance, necessity, absence of an exclusionary rule, and a properly qualified expert. White Burgess Langille Inman v. Abbott and Haliburton Co., 2015 SCC 23 added the impartiality requirement and confirmed the court's gatekeeping discretion. A forensic media authentication expert must be prepared to establish the reliability of the underlying methodology, the examiner's qualifications, and the examiner's independence from the retaining party. SWGDE-aligned methodology and documented, reproducible analytical procedures are the primary reliability indicators courts will examine.
Ontario Regulation 384/24 — Expert Report Certification (in force December 2024)
This regulation amended Ontario's Rules of Civil Procedure to require every expert report filed in civil proceedings to include a certification that the expert is satisfied as to the authenticity of every authority, document, or record cited in the report. Experts are exempt from this certification requirement for documents whose authenticity they call into question in the body of the report, documents provided by the retaining party for analysis, and documents cited solely to respond to another expert's use of the same source. For a forensic media authentication expert, the report must be structured to address this certification requirement directly.
Federal Court Practice Notice on AI — May 7, 2024 (updated June 20, 2025)
The Federal Court requires parties using AI to generate or assist in preparing filed materials to declare that use and mandates that lawyers verify the accuracy of AI-generated content before filing. The Notice explicitly identifies deepfakes as a Court concern. Potential consequences for non-compliance include adverse cost orders and contempt findings. Manitoba, Yukon, Nova Scotia, and Quebec's Court of Appeal have issued parallel AI-use notices. Ontario's Civil Rules Committee opened a 2025 consultation on proposed procedural rules specifically addressing challenges to AI-generated or AI-altered evidence.
Pre-Trial Protocol 1: Hardening Your Own Authentic Evidence
If digital media is central to your client's case and you have reason to expect an opposing-party deepfake challenge, complete these steps before documentary disclosure.
Document the capture device. Obtain and preserve the original recording device. Document the device make, model, operating system version, and any relevant configuration at the time of capture. This produces the foundational corroboration for a section 31.1 authentication argument.
Preserve the original file with its metadata intact. Do not open the original file through applications that modify access timestamps. Generate a hash value of the original file immediately and record it. The hash value is the integrity anchor: if the file is challenged as having been modified after capture, the original hash value proves it was not.
Obtain a Tier 1 screening before disclosure. A preliminary forensic authentication that confirms metadata consistency, encoding profile consistency, and the absence of detected manipulation indicators gives you a documented basis for your section 31.1 authentication argument and pre-empts a bare deepfake allegation before it is raised.
Consider C2PA content credentials where available. The Coalition for Content Provenance and Authenticity C2PA v2.2 standard (May 2025) embeds a cryptographically signed provenance chain directly in the media file at the moment of capture. A growing number of camera manufacturers are implementing C2PA at the hardware level. For prospective evidence capture where circumstances allow, a C2PA-credentialed recording provides the strongest available provenance chain.
Brief your client on chain of custody before transmission. A file shared through a messaging application, compressed, screenshotted, or re-saved has a damaged chain of custody. Brief clients on evidence-preservation obligations at the earliest opportunity.
Pre-Trial Protocol 2: Responding to a Deepfake Challenge Against Your Evidence
If opposing counsel raises a deepfake challenge against your client's authentic media after disclosure, the procedural response depends on the stage of proceedings and the seriousness of the challenge.
A bare deepfake allegation without forensic foundation does not meet the threshold required by Mohan and the weight of persuasive authority from Huang v. Tesla, United States v. Reffitt, and Wisconsin v. Rittenhouse. Your first response is to require opposing counsel to identify the specific basis for the challenge. If they cannot produce a forensic basis, the allegation is unfounded.
If a forensically grounded challenge is raised, the response is a Tier 2 comprehensive authentication analysis. This produces a documented, SWGDE-aligned forensic report that addresses the specific technical allegations raised by the opposing party and produces a conclusion with stated limitations that can withstand cross-examination.
In Ontario civil proceedings, Ontario Regulation 384/24 governs the form of the expert report. Ensure the authentication report includes the required certification statement and is served in accordance with the expert report rules under Rule 53.03 of the Rules of Civil Procedure.
At the pre-trial conference or case management stage, raise the deepfake allegation as a preliminary issue requiring resolution before trial. Do not allow an unanswered authenticity challenge to sit on the record.
Pre-Trial Protocol 3: Challenging Opposing-Party Media You Believe Is Fabricated
If you have reason to believe that media tendered by the opposing party has been AI-generated or manipulated, the challenge must be forensically grounded to succeed and must be raised at the appropriate stage.
Request the original file at the discovery stage. Do not accept a copy, a screenshot, or a compressed version. Under Ontario's Rules of Civil Procedure, a party is entitled to request production of documents in their original electronic form with metadata intact. A party who refuses to produce the original file where authenticity is in dispute invites an adverse inference.
Commission a Tier 1 screening of the produced file before raising a formal challenge. A preliminary assessment that identifies specific indicators of manipulation, metadata-hardware mismatches, encoding inconsistencies, or compression artifact anomalies gives you the forensic basis that courts require before taking an authenticity challenge seriously.
If the Tier 1 screen identifies indicators, commission a Tier 2 comprehensive analysis. Document findings at the level required for expert evidence under Mohan and White Burgess.
Raise the challenge at the earliest available pre-trial proceeding, not at trial. A mid-trial deepfake allegation without prior disclosure of the forensic basis will be treated as tactical delay. The Ontario Civil Rules Committee's proposed AI evidence challenge mechanism, under consultation since 2025, will likely formalize a pre-trial challenge pathway.
Serve the authentication expert report in accordance with Rule 53.03 and ensure it meets the Ontario Regulation 384/24 certification requirements.
Conduct Considerations
Two conduct issues arise specifically in the deepfake defense context that do not arise in conventional evidence disputes.
Raising a deepfake challenge without forensic basis. A deepfake allegation without forensic foundation raised in bad faith to delay proceedings, increase costs, or cast unwarranted doubt on authentic evidence raises professional conduct concerns under the Federation of Law Societies Model Code obligations of candour and civility. The pattern of sanctions in AI-related conduct cases in Canadian courts, including Ko v. Li (2025 ONSC 2766), Hussein v. Canada (2025 FC 1060), and Zhang v. Chen (2024 BCSC 285), demonstrates that Canadian courts are willing to impose cost consequences for AI misuse in litigation. A meritless deepfake challenge carries the same risk.
Tendering evidence without authenticating it against a known deepfake risk. Where a party knows, or ought to know, that their client's media evidence may be challenged on deepfake grounds, failure to authenticate it before tendering raises questions of competence under the Model Code. A litigator who tenders a dashcam recording in a personal injury matter without having confirmed that no manipulation indicators are present, in a file environment where the opposing party has signaled intention to challenge the evidence, is not adequately managing the forensic risk of the file.
Tuesday 14 April 2026
Authenticating evidence in the age of AI
Christopher Naudie
Increasing AI in litigation: advantages, perils and increasing regulation
Amendments to the Rules of Civil Procedure (O. Reg 384/24) - Onatrio Bar Association
Frequently Asked Questions
What is the Liar's Dividend and how does it apply in Canadian civil proceedings?
The Liar's Dividend is a concept developed by legal scholars Chesney and Citron in 2019 to describe the unintended consequence of widespread deepfake awareness: once the public knows that realistic fabricated media is possible, anyone can claim that genuine evidence is synthetic. Under section 31.1 of the Canada Evidence Act, the proponent of digital evidence carries the authentication burden. An opportunistic deepfake objection in a Canadian civil case shifts cost, time, and proof burdens onto the party with the legitimate evidence without requiring the objecting party to produce any proof of fabrication. The tactic succeeds procedurally, by increasing costs and uncertainty, more often than it succeeds on the merits.
Does the deepfake defense succeed in Canadian courts?
No Canadian appellate decision has accepted a bare deepfake defense. R. v. Medow acknowledged the deepfake threat but declined to infer manipulation without supporting evidence. Justice Nordheimer in R. v. Aslami identified the authentication risk but did not alter the section 31.1 burden. Persuasive global precedent from Huang v. Tesla and United States v. Reffitt establishes that courts will reject forensically unsupported deepfake challenges.
What does Ontario Regulation 384/24 require of expert reports in civil proceedings?
Every expert report filed in Ontario civil proceedings since December 1, 2024 must include a certification that the expert is satisfied as to the authenticity of every authority, document, or record cited in the report. Exceptions apply for documents the expert calls into question in the report, documents provided by the retaining party for analysis, and documents cited solely to address another expert's use of the same source. A SWGDE-aligned forensic authentication report produced to these requirements meets the certification standard directly.
When should a litigator commission forensic authentication rather than relying on lay authentication?
Any file where authenticity is likely to be contested, where the media was not captured under controlled conditions with documented chain of custody, or where the opposing party has indicated intent to challenge the evidence warrants forensic authentication. A Tier 1 screening returned within 48 to 72 hours is a proportionate first step in most civil files. Tier 2 comprehensive analysis is warranted where the Tier 1 screen identifies indicators or where the file will be a central exhibit at trial.
Does forensic authentication guarantee that fabricated evidence will be detected?
No authentication methodology guarantees detection of all fabrications. SWGDE methodology requires that every forensic report state its limitations clearly. What forensic authentication provides is a documented, reproducible examination that identifies the technical indicators present in the file, supports or challenges the authenticity claim to the degree the evidence permits, and is transparent about what it cannot conclude. This is the standard courts require of expert evidence under Mohan and White Burgess.
