Professor Benjamin Perrin’s presentation explored the rapidly evolving role of artificial intelligence (AI) in policing, emphasizing both it’s potential and significant risks. He began by acknowledging that no single group currently holds all the expertise needed to guide AI adoption effectively. Instead, he advocated for interdisciplinary collaboration among legal experts, technologists, police practitioners, oversight bodies, and privacy commissioners to ensure responsible implementation in high-stakes public safety contexts.

A key concern raised throughout the presentation was the lack of independent, scientific validation of many AI tools currently being used or considered by police agencies. While some technologies have demonstrated improvements in efficiency or investigative capacity under controlled conditions, evidence of their real-world accuracy, reliability, and effectiveness remains limited. At the same time, Perrin highlighted widely documented risks including bias, errors, and privacy intrusions – many of which are not yet fully mitigated. The absence of clear, policing-specific legal frameworks and the lack of Charter-based judicial scrutiny further compound these concerns.

Professor Perrin provided practical examples to illustrate these challenges, including facial recognition technologies, automated license plate readers, and AI-assisted report-writing tools such as AXON’s Draft One, which is currently being piloted in Canada. He discussed the “verification-value paradox,” where AI-generated outputs may save time but require extensive review to ensure accuracy. Issues such as automation bias, where users assume AI outputs are correct, along with documented errors (including fabricated details or “hallucinations”), raise serious reliability and disclosure concerns. He also referenced real-world cases, such as State v. Carr (2024) and R v. J.L. (2000 SCC 51), underscoring the legal scrutiny applied to novel forms of evidence.

The presentation also examined significant privacy and ethical implications. Perrin pointed to the Clearview AI case in Canada, where billions of images were scraped without consent to create a facial recognition database – an action found to violate privacy laws. He stressed the importance of understanding how AI systems are trained, including potential biases in datasets, and questioned the legal authority underpinning the use of various databases in investigative contexts. These concerns highlight the need for transparency, accountability, and clear policy guidance.

To address these challenges, Perrin outlined emerging governance frameworks and best practices. He emphasized the importance of internal oversight mechanisms such as those being developed by the RCMP, Vancouver Police Department, and others, as well as the need for meaningful third-party oversight. Tools like Privacy Impact Assessments, Human Rights AI Impact Assessments, and transparency frameworks such as the RCMP’s National Technology Onboarding Program were highlighted as key components of responsible adoption.

In closing, Perrin stressed the importance of a precautionary, self-governed approach to AI integration in policing, supported by strong accountability measures and cross-sector collaboration. He encouraged agencies to remain focused on maintaining public trust while navigating this complex and rapidly evolving technological landscape.

Professor Benjamin Perrin’s research and case materials can be found at: https://benjaminperrin.ca/ai,