Selling AI to the Police: The Digital Surveillance Business

At the recent technology conference of the International Association of Chiefs of Police (IACP) in Fort Worth, Texas, what was called 'the future of policing in the digital age' was presented. Startups and tech giants showcased artificial intelligence systems designed to automate critical police tasks, such as report writing, evidence analysis, and predictive surveillance. However, the event was marked by controversy: the press was not allowed in, and attendees described an environment where the promise of efficiency clashes with concerns about algorithmic bias, privacy, and due process.

The Rise of Police AI in Fort Worth: Historical and Technological Context

The 2025 IACP conference represents a milestone in the growing integration of artificial intelligence into policing, a trend dating back to the 2010s with the early adoption of facial recognition and predictive analysis systems. Companies like Axon, known for its Taser body cameras, have expanded their offerings to AI platforms like Axon AI, which automatically transcribes camera recordings and generates preliminary reports. Palantir, famous for its work with intelligence agencies, offers its Gotham platform to integrate data from multiple sources and generate real-time alerts. ShotSpotter, now part of SoundThinking, uses acoustic sensors to detect gunshots and deploy police resources more quickly. According to The Verge, these technologies promise to reduce the administrative burden on officers, but also pose risks of discrimination and errors if not properly supervised. The exclusion of the press from the event has drawn criticism, as it limits transparency about the actual capabilities and potential biases of these systems.

What Exactly Is Being Sold? Details of the Tools

The systems presented include several key categories: facial recognition, real-time video analysis, crime prediction algorithms, and report automation. Facial recognition, offered by companies like Clearview AI and NEC, allows identifying suspects from surveillance camera images or social media. However, a 2019 study by the National Institute of Standards and Technology (NIST) found that many algorithms have significantly higher error rates for Black and Asian individuals, with false positives up to 100 times more frequent in some cases. Real-time video analysis, such as BriefCam's system, summarizes hours of footage in minutes and detects suspicious objects or behaviors. Crime prediction algorithms, like PredPol (now part of SoundThinking), use historical data to forecast areas with higher likelihood of incidents, but research such as that by the RAND Corporation has questioned their effectiveness, noting they often simply reinforce existing surveillance patterns without reducing crime. Report automation, exemplified by products like Axon's, uses natural language processing to draft legal documents from transcriptions, which could save hours of work but also introduces risks of factual errors or linguistic biases. According to Wired, these systems are often trained on historical data that may contain racial biases, perpetuating inequalities.

Impact on the Justice System: Acceleration vs. Bias

Automating critical legal steps, such as drafting incident reports or identifying suspects, could speed up processes but also introduce bias. For example, a 2021 ACLU report documented cases where facial recognition led to wrongful arrests of innocent people, such as the case of Robert Williams in Detroit, who was incorrectly detained due to a false positive. Moreover, the lack of transparency in these systems hinders accountability: proprietary algorithms are not publicly auditable, and police agencies often sign confidentiality agreements with vendors. This contrasts with earlier events, like the adoption of video surveillance in the 1990s, which also sparked privacy debates but allowed for greater judicial oversight. Police AI introduces an unprecedented level of opacity, as algorithmic decisions can be difficult to challenge in court. A 2022 Stanford University study found that prosecutors increasingly use AI-generated reports as evidence without questioning their accuracy or potential biases.

Consequences for Society: Public Trust and Predictive Surveillance

While AI can improve police efficiency, its implementation without robust legal frameworks could erode public trust. Civil rights organizations, such as the Electronic Frontier Foundation, have pointed out that predictive surveillance can lead to overrepresentation of certain communities, creating a vicious cycle of surveillance and criminalization. For instance, in Los Angeles, the use of PredPol in 2018 resulted in an increase in police stops in minority neighborhoods without a corresponding reduction in crime rates. The sale of these technologies represents a multi-billion dollar business: according to MarketsandMarkets, the AI surveillance market reached $12 billion in 2025, with 20% annual growth. However, federal regulation in the United States remains fragmented. While cities like San Francisco and Boston have banned facial recognition, others like Fort Worth have adopted it without restrictions. Industry lobbying is strong: according to OpenSecrets, police technology companies spent over $50 million on lobbying in 2024. This contrasts with the European Union, where the 2024 AI Act classifies live facial recognition as high-risk and requires conformity assessments.

What Readers Should Know: Actions and Perspectives

It is crucial for citizens to demand transparency about how these tools are used. Some cities have already implemented independent audits, such as the Seattle Police Office of Inspector General, which reviews AI systems annually. Readers should inform themselves about their local police department's policies and advocate for legal frameworks that include human review of all automated decisions. Additionally, public education on algorithmic bias is essential: organizations like the Algorithmic Justice League offer resources to understand how these systems work. Finally, political pressure can make a difference: in 2024, a citizen petition in Austin, Texas, led the city council to pass an ordinance requiring prior approval for any surveillance technology. The future of policing in the digital age is not written; it depends on active societal participation to ensure that AI serves justice, not oppression.