Despite growth, many departments remain cautious about integrating artificial intelligence (AI) tools into policing practices
A new survey by the University of Michigan reveals a steady increase in the use of surveillance technology by local law enforcement across the state over the past nine years. However, despite this growth, many departments remain cautious about integrating artificial intelligence (AI) tools into policing practices.
The findings come from the Michigan Public Policy Survey (MPPS), a long-running study conducted by the Center for Local, State, and Urban Policy at the University of Michigan. Since 2009, the survey has gathered insights from 1,856 local governments, including their leaders, law enforcement agencies, and elected prosecutors.
The spring 2024 edition of the survey focused on the adoption of surveillance and public safety technologies such as body and dashboard cameras, drones, facial recognition, and automated license plate readers. It also explored perceptions around the role and trustworthiness of AI in law enforcement and judicial decision-making.
The results show a marked increase in surveillance equipment usage. In 2015, only 23 per cent of law enforcement agencies reported using body-worn cameras. That figure has now risen to 77 per cent. Similarly, dashboard cameras saw an increase from 55 per cent to 74 per cent over the same period. Public surveillance camera use has grown from 28 per cent to 50 per cent, while drone usage jumped from just 3 per cent in 2015 to 31 per cent in 2024.
While these technologies have been widely adopted, the use of AI and machine learning in criminal justice remains a topic of hesitation. When asked about the reliability of automated decision-making tools compared to human judgement, 55 per cent of local government officials said they were unsure. That uncertainty rose among county sheriffs and police chiefs (59 per cent), and was even higher among county prosecutors (66 per cent).
Prosecutors were also surveyed about their trust in specific applications of AI. Half of the respondents expressed at least some trust in AI systems that identify high-risk neighbourhoods, while 45 per cent said the same for AI tools used to analyse forensic evidence. However, only 20 per cent indicated any level of trust in AI tools used for risk assessments related to sentencing, parole, probation, or release decisions.
The study’s authors point out that the increased reliance on surveillance technologies has prompted a timely discussion about their value, as well as the potential and limitations of AI. Recent investments further illustrate the trend. In 2023, the Grand Rapids Police Department allocated nearly USD 100,000 to expand its drone programme. That same year, Michigan State University made a significant investment in facial recognition technology for its surveillance network, following a campus shooting that resulted in three student deaths and several injuries.
As AI continues to evolve and its use in public safety grows, the researchers suggest that it is essential for local governments and law enforcement agencies to evaluate both its capabilities and its risks. The MPPS findings indicate that while technological surveillance has become more commonplace, the path towards adopting AI in critical legal and law enforcement decisions will likely remain cautious and gradual.
