How the threat surface for police evidence is shifting, and where it isn’t
AI is the question Canadian police chiefs and IT directors are being asked about most often right now, and it is the question most are least prepared to answer. The questions come from city councils, from procurement officers, from journalists, from rank-and-file officers wondering what tools they should and should not trust. The pressure to have a coherent answer is real, and growing.
For evidence specifically, the concern is well-founded. Generative AI can fabricate audio, edit video, alter dash-cam footage, and produce images convincing enough to deceive the human eye. Dash-cam footage may be altered to make a red light look green. Defence counsel knows this. Crown counsel knows this. Police evidence handlers know this. Across Canadian courts, the question of what counts as reliable digital evidence is being renegotiated in real time, case by case, judge by judge.
But not all police evidence is equally exposed to the AI question. Some forms of evidence sit directly in the path of AI manipulation. Others, despite the panic, sit largely outside it. The distinction matters for any Canadian agency thinking about which digital tools to invest in over the next procurement cycle.
Visual evidence is the front line
Photographs, video, and dash-cam footage are where the AI threat lands hardest. The tools to alter them are accessible, the alterations are increasingly difficult to detect with the human eye, and the legal challenge to authenticity is now a standard tactic. Defence counsel may raise the possibility that a video was manipulated or altered by AI. Although simply raising the possibility of manipulation is speculative, the trial judge may still have a reasonable doubt.
This is not entirely new. Photographic manipulation predates AI by decades. What AI changes is the speed, accessibility, and quality of the manipulation, not the underlying problem.
For Canadian agencies, the implication is that authentication infrastructure (chain of custody, cryptographic sealing, tamper-evident storage) is now table stakes for visual evidence. The agencies that built for this five or ten years ago are in a different position than the agencies that did not.
Officer notes sit on the other side of the line
Officer notes are a different category. They are textual records, not visual ones. They are either locked or they are alterable. They are either tamper-evident or they are not. The integrity question does not turn on whether an AI tool can convincingly forge them. It turns on whether the platform that holds them allows for alteration in the first place.
That distinction is one that prosecutors who have argued the integrity of digital notes in Canadian courts recognize quickly. Hans Saamen, recently retired Director of Prosecutions for the Regional Municipality of York, points to his own pre-AI experience: photographs of scenes were already being doctored to fabricate evidence. In one case from his career, a stop sign was Photoshopped into a different location to make it appear that a tree was blocking the stop sign as they approached the intersection. In another case, the stop sign was actually removed from the location to argue it was not there. The methods existed before AI made them easier. As Mr. Saamen puts it:
“In terms of officers’ notes, I don’t see how AI really is going to be helpful to alter officers’ notes, but the videos and photographs, yes.”
The standard for officer notes did not change because AI arrived. It was clear before AI, and it is clear now. What has not been clear is whether Canadian agencies have built for it.
What courts actually test
Part of the gap comes from a common misunderstanding about what an officer’s notes are, in court.
An officer’s testimony is the evidence at trial. The notes are not. The notes function as a memory aid. The officer takes the stand, refers to the notes to refresh recollection, and gives testimony based on what they observed and recorded. The notes themselves are not entered into evidence unless their authenticity is challenged.
That technicality has substantive consequences. When defence counsel sets out to attack an officer’s testimony, the most effective line of attack is the integrity of the notes on which the testimony is built. The lawyer does not need to prove the notes were altered. Raising the possibility that they could have been altered is enough to plant doubt, and doubt is enough to undermine the testimony.
This is where the actual reliability question lives. Not in whether AI can forge a notebook entry, but in whether the platform an agency has chosen produces records that cannot be altered after the fact, and whether the officer can demonstrate that fact under cross-examination.
Where Canadian agencies are most underprepared
The technology gap on this question is smaller than the training gap. Many digital eNotes platforms in Canadian agencies do produce locked, tamper-evident records. What officers struggle with is being able to explain and demonstrate those features when their testimony is challenged.
This is a pattern that recurs in mock cross-examination exercises, and Mr. Saamen has seen it first-hand:
“Officers knew that the notes were secured and were locked and couldn’t be altered, but they couldn’t explain the security features and they couldn’t demonstrate it. That was the problem.”
A platform that is secure but unexplainable is not, in courtroom terms, secure enough. An officer who can describe the security model, log into the system on the stand, and show the original locked record while the alleged altered version is on the screen is in a different position from an officer who cannot. The first walks out of cross-examination with their testimony intact. The second does not.
Closing this gap is, increasingly, where Canadian agencies should be focusing procurement attention. Not “is this platform secure?” but “can our officers demonstrate that it is?”
Three questions every Canadian agency should ask a software vendor
For any procurement evaluation of a digital officer notes platform, three questions belong at the top of the list:
- Once a note is locked, can it be altered by anyone, including the officer who wrote it? The answer must be no.
- What security features prove the note has not been altered, and can an officer demonstrate those features on the witness stand? The answer must include features the officer can show in real time, not just internal logs no one will see in court.
- What training does the vendor provide so officers can explain and demonstrate the security model under cross-examination? A platform that is secure but cannot be explained is not, in practical terms, court-ready.
How DMI builds for this
DMI’s eNotes platform is built around exactly this standard.
Once an officer locks a note, it cannot be altered by anyone, including the officer who created it. The platform is protected by Braidmark®, DMI’s patented document security technology, which produces a cryptographic seal an officer can demonstrate in court. The system records who created each note, when, where, and every interaction with it, producing an auditable trail that holds up to scrutiny.
DMI also works with deploying agencies on training. Officers receive instruction not only on how to use the platform, but on how to explain and demonstrate its security features when their testimony is tested.
This is what we mean when we say DMI’s products are designed for Canadian policing realities. The bar is not “digital.” The bar is “court-proof.”
Conclusion
AI is reshaping the conversation about police evidence in Canada, and Canadian agencies are right to take the shift seriously. For photographs, video, and dash-cam footage, the threat surface is genuinely changing, and the authentication infrastructure agencies invested in over the past decade is being tested in new ways.
For officer notes, the picture is different. The standard that holds up in court has not changed. Locked records that cannot be altered after the fact, security features the officer can demonstrate from the witness stand, and training that gives the officer the language to do it. That standard existed before AI was a serious factor in any of these conversations, and it will exist after the current wave of attention passes.
What has changed is how visible the gap is between the Canadian agencies that have built for this standard and the agencies that have not.
Frequently asked questions
Does AI affect the reliability of officer notes in Canadian courts?
Officer notes are textual records that are either locked or alterable. AI does not meaningfully change that question. The threat to officer note integrity is whether the platform that holds them allows alteration in the first place. The threat to visual evidence (photographs, video, dash-cam) is more directly reshaped by AI.
Are officer notes considered evidence in Canadian courts?
No, the officer’s testimony is the evidence. Officer notes function as a memory aid that the officer refers to on the witness stand. The notes themselves are not entered into evidence unless their authenticity is challenged.
What makes a digital officer notes platform “court-ready”?
Three things: notes that cannot be altered by anyone (including the officer who wrote them) once locked, security features the officer can demonstrate in real time under cross-examination, and training that gives the officer the language to do it.
What is the AI threat to police evidence?
The threat lands hardest on visual evidence: photographs, video, dash-cam footage. Generative AI can fabricate or alter these convincingly. Authentication infrastructure (chain of custody, cryptographic sealing, tamper-evident storage) is increasingly essential for visual evidence. For textual records like officer notes, the integrity question is structural rather than AI-specific.
Source note
Includes commentary from Hans Saamen, B.A., LL.B., recently retired Director of Prosecutions for the Regional Municipality of York and a Lifetime Member of the Prosecutors’ Association of Ontario, drawn from a May 25, 2026 interview.
