Toronto Police to Implement AI for Non-Emergency Calls

Toronto Police Service (TPS) is set to introduce artificial intelligence (AI) technology to manage non-emergency calls beginning in February 2024. This initiative, announced during a recent board meeting, aims to improve response times for callers seeking assistance for non-urgent matters. According to TPS Chief Transformation Officer Colin Stairs, the integration of AI is part of a broader strategy to enhance service efficiency.

Approximately one third of the 1.9 million calls received annually by TPS fall into the non-emergency category. In 2025, residents of Toronto experienced an average wait time of six minutes and 11 seconds for such calls. This new technology, developed by Hyper, a Canadian AI voice call startup, is expected to significantly reduce these wait times by streamlining the call-handling process.

How AI Will Operate

The AI system will engage with callers to confirm essential details, such as their location and whether someone has been injured, before providing appropriate instructions. Hyper has indicated that the AI tool is trained using data from actual 911 calls, which informs its responses and question sequences. A demonstration reported by The Globe and Mail showcased the AI’s capabilities in a mock scenario. Damian McCabe, co-founder of Hyper, illustrated how the AI interacted with him after he described being involved in a minor car accident. The AI agent posed relevant questions while generating typing sounds to encourage concise responses from the caller.

In a positive development, Winnipeg Police reported that the implementation of Hyper’s technology resulted in a 10% decrease in call hang-ups, indicating a favorable public reception of the AI system.

Challenges and Concerns

Despite the promising advancements, there have been instances highlighting the limitations of AI in policing contexts. For example, Halton Regional Police Service fully integrated Hyper’s technology in October 2023. During a call, the AI incorrectly advised a caller that a minor incident did not require reporting, based on the fact that the caller was not in the vehicle at the time. A human operator later explained that the incident should have been reported, underscoring the complexities involved in programming AI to interpret nuanced situations.

Moreover, other police departments have faced challenges stemming from AI misuse. Recently, an incident involving the chief of the West Midlands police in England revealed that fabricated evidence produced by an AI tool misled officials regarding crowd control measures. Additionally, a bizarre AI-generated police report in Utah claimed that an officer had transformed into a frog during a routine traffic stop, highlighting the unpredictable nature of AI in law enforcement.

The forthcoming deployment of Hyper’s AI platform for TPS will specifically assist in handling non-emergency calls through the 811 service. However, TPS has assured the public that emergency calls will continue to be managed by human operators, maintaining a critical human element in urgent situations. As TPS prepares for the AI rollout next month, the effectiveness and reliability of this technology remain to be fully assessed.