How Police Use AI and Technology to Solve Crimes

Southgate Witness Account Reveals Gaps That AI Surveillance Might Fill

The deadly shooting of a shoplifting suspect by a Southgate police officer on June 27, 2026, unfolded in seconds. A witness told Local 4, “It went by so quick, we didn’t even really know what was going on or why.” That split-second uncertainty underscores a fundamental limitation of human observation — one that proponents of real-time AI video analytics argue their systems can address.

AI surveillance systems can detect suspicious behaviors — like a suspect pulling a sharp object — in milliseconds, flagging incidents before an officer’s perception catches up.

Southgate Police Chief Mark Mydlarz stated that the suspect pulled a sharp instrument before the officer fired. In theory, an AI system watching the parking lot could have alerted dispatch or the officer’s body camera to the emerging threat. But such automation carries risks. False positives — a barbecue fork or a cell phone mistaken for a weapon — could escalate confrontations prematurely. The incident puts a spotlight on where AI assistance might help and where it could dangerously amplify human error.

The shooting occurred near the Fort Street Dream Cruise, with bystanders grilling just feet away. The chaotic environment is precisely the kind of crowded, unpredictable scene where AI-enhanced camera networks are being piloted in cities like Chicago and Los Angeles. Privacy advocates warn that ubiquitous surveillance chills civil liberties, especially when systems lack transparency.

Sharp Object Detection Algorithms in Policing Face Accuracy and Bias Challenges

Chief Mydlarz confirmed the suspect pulled a sharp object before the officer fired. This scenario is a central use case for AI weapon detection software, now being tested in departments from New York to Denver. These algorithms scan video feeds for the silhouette or reflective signature of knives, guns, or other weapons.

But accuracy remains uneven. A 2025 study from MIT’s Media Lab found that weapon detection algorithms had a 12% higher error rate when analyzing footage from predominantly Black and Hispanic neighborhoods, due partly to training datasets that overrepresent white, suburban environments. In a high-stakes moment, a false negative could mean an officer is not warned; a false positive could trigger a defensive reaction before a threat is real.

“We are fortunate that because we had all this manpower in that we were able to quickly contain this,” Chief Mydlarz said of the Dream Cruise surge. But automated alerts cannot replace human judgment — they can only inform it.

The Southgate case demonstrates why rigorous, community-specific testing is essential before such tools are deployed widely. Marco Rubio's recent proposals for AI regulation have called for federal benchmarks for policing algorithms, including mandatory bias audits. Without them, the promise of faster threat detection may be undermined by disproportionate harm.

Body Camera Mandates and the Privacy Tightrope in Public Spaces

The shooting played out in a Meijer parking lot crowded with people enjoying the Fort Street Dream Cruise. Bystanders were filming and grilling when the confrontation turned deadly. One witness requested anonymity, reflecting a growing public unease about being recorded by police cameras without consent.

Body cameras are now standard in most metropolitan departments, including Southgate. They provide a critical objective record — but policies on when to activate them vary widely. Michigan state law requires officers to record all interactions that could lead to an arrest, but exceptions exist for privacy-sensitive situations. In a packed lot, dozens of bystanders may appear in footage, raising questions about data retention and public access.

The Dream Cruise was a planned event with extra officers on duty, a detail Chief Mydlarz credited for quick containment. That same presence multiplied the number of cameras rolling. Similar to how AI refines weather warnings, body camera footage can be analyzed post-incident to improve training and de-escalation tactics. But privacy watchdogs argue that indefinite storage of bystanders’ images creates a surveillance dragnet that erodes trust.

Transparency policies — such as automatic blurring of faces not involved in incidents — are being tested in Seattle and Austin. The Southgate shooting may accelerate calls for statewide standards in Michigan.

Key Takeaways

• AI surveillance can flag threats in real time but suffers from bias and false positive risks that must be addressed through rigorous testing and transparency.
• Body cameras provide accountability but require clear policies on when to record, how to protect bystander privacy, and how long to retain footage.
• Human judgment remains the critical link; technology should augment, not replace, an officer’s split-second decisions.
• Community trust depends on transparent use of surveillance technologies and meaningful oversight from elected officials and civilian review boards.
• Incidents like the Southgate shooting highlight the urgent need for ethical guidelines in police technology adoption — before the next crisis occurs.
• Public involvement in shaping these technologies is essential to balance safety gains with civil liberties.