Innocent Driver Detained After AI Cameras Misread Number Plate

An blameless automotive journalist was surrounded by armed police officers in a Minnesota car park after AI-powered surveillance cameras misinterpreted his vehicle’s number plate and wrongly flagged it as stolen. Joel Feder was testing a £155,000 Range Rover near Plymouth in late June when four police cars boxed him into a Kohl’s car park, with officers stepping out with hands positioned near their weapons and demanding he exit the vehicle. The event occurred after Flock Safety number plate cameras had been monitoring Feder’s Range Rover for several days, continually misreading the New Jersey manufacturer registration plate due to its smaller central digits. Officers ultimately confirmed the vehicle was genuine after contacting Jaguar Land Rover, discovering the allegedly stolen registration had merely been misplaced during a photo shoot.

The Misidentification That Resulted in Armed Police Response

The underlying cause of the incident lay in a simple yet devastating failure of the AI system to correctly identify Feder’s number plate. His Range Rover bore a New Jersey manufacturer plate reading 34 10 DTM, with the number 10 purposefully displayed smaller than the surrounding characters—a typical approach by Jaguar Land Rover. When a dealership in Los Angeles flagged a different plate, 34 03 DTM, as missing, the police report mysteriously recorded it only as 34 DTM, leaving out the middle numbers entirely. Flock’s cameras, unable to register the smaller digits on Feder’s plate, repeatedly read it as 34 DTM, producing what appeared to be a perfect match to the flagged vehicle.

What made matters worse was the compounding effect of the error. Once the AI system had identified Feder’s vehicle, it continued tracking him around Plymouth for several days, constructing a false case against him. Police officers, acting on what they believed to be reliable technological evidence, stationed themselves to intercept the Range Rover. When Feder’s vehicle entered the Kohl’s car park, officers intervened promptly, treating the situation as a legitimate vehicle retrieval rather than scrutinising the AI system’s accuracy. The incident exposed a concerning shortfall in protocol: officers had seemingly accepted the AI’s identification without adequate human review in advance, prioritising swift action over thorough examination.

• Flock Safety cameras were unable to identify lower-positioned numbers on plate
• Police report documented stolen plate without complete number sequence
• System kept monitoring Feder’s vehicle for multiple days unquestioned
• Three additional similar vehicles were under surveillance in Minnesota simultaneously

How Machine learning systems Was unable to Interpret the Plate Properly

The Technical Glitch Underlying the Mix-up

The core problem resulted from Flock Safety’s lack of capacity to correctly interpret the unique structure of Jaguar Land Rover’s manufacturer plates. These plates display a reduced centre digit—in Feder’s case, the “10” in “34 10 DTM”—a styling choice that people easily identify but which stumped the AI system. When the imaging systems captured his vehicle, they just failed to capture these diminished figures, instead processing the plate as “34 DTM”. This excessive simplification created an exact match with the partial police report for the missing plate, triggering the false positive that would lead to his detention.

The error exposes a fundamental flaw in automated vehicle identification technology: its reliance upon standardised formatting assumptions. Flock’s system appeared programmed to require standardized character sizing across all plates, rendering it unable for the lawful variations that manufacturers implement. Rather than flagging the discrepancy for human verification, the AI proceeded with confidence with its faulty analysis, processing the abbreviated plate number as trustworthy information. This confidence in an incorrect result proved dangerous, as officers downstream had little cause to question the automated data they’d been provided.

What created this failure notably distressing was its ripple impact through the whole enforcement system. Once Flock’s cameras had misidentified Feder’s vehicle, the system intensified the problem by continuing to track him across various days and places. Each additional sighting reinforced the false match, constructing an progressively persuasive but entirely fabricated case against an uninvolved driver. Police officers, trusting the technology’s output, allocated resources to apprehend what they considered to be a stolen vehicle. The incident demonstrates how AI errors don’t just fade away—they compound and escalate, possibly converting a small technical fault into a significant police matter with actual repercussions for innocent people.

• AI system failed to identify smaller-sized middle characters on plate
• Incomplete police report excluded complete digit sequence from vehicle theft
• System persisted in monitoring without human verification or accuracy verification

The Mounting Challenge of AI Errors in Police Operations

Feder’s experience is anything but unique. The incident underscores a troubling pattern of AI technology making catastrophic mistakes across police departments throughout America. These aren’t minor information handling mistakes that disappear quietly in a system’s logs—they’re failures that directly endanger law-abiding citizens, leading to armed law enforcement action, incarceration, and emotional harm. As artificial intelligence grows more integrated in policing infrastructure, from licence plate recognition to facial recognition and behaviour assessment, the consequences of such failures have reached unprecedented levels. Each malfunction constitutes a breakdown in the safeguards meant to protect people from unjust detention and harassment.

The troubling reality is that many of these systems function with limited human supervision or verification mechanisms. Police departments have implemented AI solutions with considerable enthusiasm, often prioritising rapid processing and effectiveness over correctness and answerability. When technology identifies a vehicle or person as suspicious, officers regularly treat that designation as essentially beyond question rather than a starting point for investigation. This unquestioning confidence in algorithms has produced a harmful cycle where technological errors are intensified through law enforcement activity, transforming digital mistakes into real-world crises that can upend lives and damage public confidence in policing.

The pattern demands urgent change. Law enforcement agencies should establish mandatory human verification procedures prior to deploying armed interventions driven by AI alerts. Software developers creating surveillance systems need more robust accountability measures and rigorous evaluation across different environmental scenarios and situations. Absent urgent intervention, ordinary citizens will remain encountering unnecessary police stops, arrest, and additional harms—all as a result of systems committed mistakes that humans did not catch beforehand.

Questions Regarding Monitoring Systems and Public Safety

The situation involving Joel Feder raises uncomfortable questions about the speed at which police departments has adopted artificial intelligence monitoring without sufficient protections. Flock Safety’s cameras are currently operating across thousands of areas in the United States, handling millions of number plate images each day. Yet this Feder case demonstrates that the technology remains prone to errors, particularly when handling unconventional plate formats. The reduced font size on his NJ manufacturer’s plate—a valid alternative—proved to be the critical problem that sparked days of surveillance and an armed law enforcement response. If a straightforward visual variation can confound the technology, how many innocent drivers are under surveillance based on computational mistakes?

The broader concern goes further than any single company’s technology. Police departments have seemingly accepted AI flagging as a shortcut to investigation, treating algorithmic matches as grounds for arrest rather than information demanding further checks. When officers reached the Kohl’s car park in a heightened state of alert, they had already accepted the AI’s conclusion as established truth. Only following the detention of Feder did they take the step of verifying the vehicle’s legitimacy by consulting Jaguar Land Rover—a step that ought to have come before any armed response. This approach reverses the proper investigative process and puts innocent citizens vulnerable to unnecessary confrontation.

What Needs to Shift

Meaningful reform necessitates measures at several tiers. Technology developers need to carry out rigorous testing across diverse licence plate formats, including OEM plates with non-standard typography. Police departments require enforceable policies demanding manual confirmation before any armed deployment occurs. Surveillance companies should face legal accountability when their systems generate mistakes that lead to wrongful detention. Finally, the public is entitled to openness about camera locations and coverage and data usage practices, enabling meaningful discussion about the relationship between security surveillance and privacy protection.

• Mandate third-party evaluation of AI systems across varied number plate types prior to implementation
• Require documented manual confirmation of flagged vehicles before any law enforcement intervention
• Establish statutory responsibility for monitoring firms whose errors cause unlawful arrest
• Implement disclosure obligations about surveillance site placement and information storage procedures

Until these protective measures exist, the technology will keep producing mistakes that transform innocent people into suspects. Joel Feder was fortunate—he was freed in under an hour and faced no charges. Others have not been so lucky. A Tennessee grandmother spent nearly six months in prison based on faulty facial recognition. The pattern indicates that without urgent intervention, more innocent people will bear the burden of algorithmic errors that police fail to detect.