AI traffic cameras in the US: they don't just read license plates, they build life profiles of citizens

By Motor1.com · June 24, 2026.

A viral post on social media has reopened the debate over mass surveillance in the United States, prompted by the proliferation of AI-equipped traffic cameras. Content creator Allie Voss (@allie_voss) warned on social media about the real capabilities of these devices, which go far beyond reading license plates or detecting speeding. Her message resonated strongly because it pointed to something concrete: the company Flock Safety, whose cameras have been installed in numerous small U.S. cities, would be building what she calls 'pattern-of-life profiles' on anyone who appears in the field of view of its cameras.

Voss explains that ALPR (Automatic License Plate Readers) systems have evolved radically. 'People are used to those typical license plate cameras used to charge tolls or catch you speeding', she notes, 'but modern ALPRs, which are basically AI-powered traffic cameras, are becoming a mechanism of mass government surveillance over things that have nothing to do with driving'. The key to the qualitative leap lies precisely in AI: where before there was a camera that read a plate and queried a database, now there are systems capable of cross-referencing information, detecting movement patterns, identifying vehicles, pedestrians and cyclists, and building a detailed history of a person's habitual movements.

According to Voss, Flock Safety's cameras are not limited to recording traffic violations. She claims they are being used to verify whether children attend the school district they are assigned to, to handle complaints about loud music, and to allow law enforcement officers to access Flock's databases in situations where they do not wish to obtain a legitimate court order. This last claim is especially serious from a constitutional standpoint, since it implies that data collected by a private company would be serving as an alternative route around the judicial oversight that governs police investigations. Voss adds that even potential employers would be accessing those databases to gather information about job candidates.

The Motor1 article provides additional context to these claims by drawing on other sources. In April, the outlet Device Daily already warned that AI-powered urban cameras 'are raising privacy alarms', and described how the technology serves to create 'a vast searchable database that can be integrated with other law enforcement data repositories'. The same text establishes a significant historical analogy: it compares these systems with the network of CCTV cameras that London law enforcement deployed in the 1970s, in the midst of the conflict with the IRA.

Device Daily also highlights that these cameras are frequently operated by private companies, not directly by public agencies. These corporations incorporate AI systems into their observation platforms, which 'significantly increases their reach'. Integration with databases such as the National Crime Information Center (NCIC) means that, if a camera records the vehicle of a suspect listed in that database, the AI can issue an instant alert to local law enforcement. The result is a real-time surveillance infrastructure that connects public space with national police records.

Flock Safety is the company most cited in this controversy. Motor1 notes that Richmond, Virginia, spent one million dollars in a single year on Flock Safety cameras. The company has been the target of criticism not only over its cost, but also over doubts about its real effectiveness in reducing violent crime, although some data do suggest that it facilitates the recovery of stolen vehicles. As for legality, Flock Safety faced a federal lawsuit in Norfolk, Virginia, alleging that images and recordings of citizens were used improperly by law enforcement. In February, a federal judge ruled that Flock's cameras do not violate citizens' liberties, though he added that this conclusion could change in the future in the face of new technological advances, a warning that proves revealing: the judiciary itself acknowledges that current law may fall short against the speed of evolution of these systems.

The phenomenon is not exclusive to the United States. Motor1 cites NBC News to frame it as a global reality. David Kelly, vice president of Acusensus, an Australian company with government contracts, argues that its system does not store images when there is no violation: 'If there is no violation, we don't keep the data. If, upon review, no penalty is issued, it isn't kept either'. However, Kelly himself acknowledges a revealing limit: he can speak for his company, but not for what local governments do with the same data. In other words, the flow of information between private company and public administration lacks a transparency framework that guarantees the data is not reused for other purposes. Acusensus also works with U.S. universities, where it tests its 'Heads Up' system, capable of detecting whether drivers are holding their phones or not wearing a seatbelt, which further broadens the spectrum of monitored behaviors.

The underlying question that frames the entire debate is that of the principle of accountability. As Device Daily notes, 'data collected through surveillance infrastructure in the U.S. can circulate with limited transparency and accountability'. Organizations such as the American Civil Liberties Union (ACLU) and the Electronic Frontier Foundation (EFF) have spent years denouncing the risks of license plate reader systems, and their warnings are taking on new urgency now that AI exponentially multiplies the capabilities of these systems. The problem is no longer whether a camera saw your car pass down a street: the problem is that an AI system can reconstruct your daily routine, know where you work, what time you usually take your children to school, and store that information in private databases with minimal public oversight.

From the perspective of agentic AI, this case clearly illustrates one of the most cited risks in regulatory debates: the ability of AI systems to aggregate seemingly innocuous data and build high-impact inferences about individuals without their knowledge or consent. A camera that records a license plate is a single data point; an AI system that correlates dozens of readings of that plate over time, at different points in a city, and cross-references them with other sources, produces something qualitatively different: a behavioral profile with predictive capacity. This leap from the descriptive to the inferential is precisely what makes agentic AI applied to surveillance fundamentally different —and potentially more invasive— than any prior surveillance technology.

At the regulatory level, the contrast with Europe is notable. As sector context, the European Artificial Intelligence Regulation (EU AI Act), in force since 2024, classifies real-time remote biometric identification systems in public spaces as 'high-risk' applications and, in most cases, expressly prohibits them except for very narrowly defined exceptions for security forces. Mass surveillance systems of the type described in this article would hardly pass European regulatory scrutiny without substantial modifications. In the United States, by contrast, regulation is fragmentary: it depends on each state and even each municipality, which creates a patchwork of highly unequal protections. Some U.S. cities have banned the use of facial recognition by the police, but AI-powered ALPRs fall into a different category that current legislation does not address systematically.

For companies and developers working in the AI ecosystem, the Flock Safety case raises important questions about accountability across the data-use chain. When a private company sells a system to a municipality, and that municipality integrates it with federal databases, and third parties —employers, other agencies— access that data, who is responsible if there is misuse? The February ruling that exonerated Flock Safety suggests that, at least for now, U.S. courts do not consider that these systems violate constitutional rights, but the judge's own warning about future technological advances indicates that this position may be provisional.

The impact on the public perception of AI is also relevant. Cases like this, amplified by social media through creators such as Voss, contribute to eroding citizens' trust in AI systems deployed in public space. The narrative of 'the camera only monitors traffic' that many municipalities have used to justify these installations without generating resistance is cracking as citizens better understand the real capabilities of the underlying technology. That gap between the official justification and the technical reality is a governance problem as much as a privacy one.

In short, the Motor1 article captures a story that has all the characteristics of a landmark case in the debate over AI and surveillance: technology deployed rapidly and with limited oversight, use that overflows the stated purposes, opacity about the fate of the data, and a citizenry that discovers after the fact the real scope of the systems surrounding it. The virality of Voss's video and the media attention it has generated suggest that the debate over the limits of AI surveillance in public space is far from settled by the February ruling.