Artificial intelligence and data-driven technologies are often described as the future of policing. From facial recognition and predictive analytics to automated surveillance tools, these systems promise greater efficiency and more strategic deployment of limited resources. Yet across the United States, a growing number of governments have chosen to slow down, restrict, or discontinue certain technologies. These decisions highlight an important lesson: adopting AI in policing is not simply a question of innovation but of governance, accountability, and public trust.
One of the most visible examples is facial recognition technology. In 2019, San Francisco became the first major U.S. city to ban government use of facial recognition, including by police departments, citing concerns about privacy, civil liberties, and accuracy. Other cities, including Boston and Portland, later enacted similar restrictions. At the state level, policymakers have pursued more targeted safeguards. California imposed a temporary moratorium on facial recognition embedded in police body cameras, while states such as Washington and Colorado implemented requirements for accountability reporting, testing standards, and oversight before agencies can deploy these systems.
Predictive policing tools have also faced growing scrutiny. The Los Angeles Police Department discontinued its PredPol and LASER predictive policing programs following audits and public criticism questioning their transparency and effectiveness. Chicago ended its “Strategic Subject List,” a data-driven system designed to identify individuals at higher risk of involvement in violence, after independent evaluations found limited evidence that the tool reduced crime and raised concerns about fairness. In Santa Cruz, California, officials went further by banning predictive policing entirely, citing worries that algorithms trained on historical data could reinforce existing patterns of enforcement rather than produce unbiased forecasts.
Legislatures have also placed limits on policing technology, introducing guardrails to ensure that technology adoption aligns with constitutional protections and public expectations. Several states now prohibit law enforcement from relying solely on facial recognition matches as the basis for arrests or probable cause, requiring independent evidence instead. Laws governing automated license plate readers often restrict data retention periods and limit how information can be shared to address privacy concerns. Drone technologies enhanced by AI capabilities have likewise been subject to warrant requirements and procurement restrictions designed to reduce cybersecurity risks.
Across these examples, three consistent concerns emerge. First, policymakers increasingly demand evidence that technologies actually improve outcomes. Programs that fail to demonstrate clear effectiveness face greater scrutiny, especially when they carry significant costs or operational risks. Second, fairness and bias remain central issues. Because many AI systems rely on historical data, there is concern that they may replicate existing enforcement patterns rather than provide neutral analysis. Third, public legitimacy plays a decisive role. Technologies perceived as opaque or intrusive can undermine community trust, even when agencies believe they improve efficiency.
These developments suggest that the future of policing technology will not follow a simple path of rapid adoption. Instead, governments appear to be moving toward a more cautious and experimental approach—one that allows innovation but treats technological deployment as reversible and subject to oversight. Decisions to limit or discontinue AI tools should therefore not be viewed as resistance to progress but as part of a broader effort to ensure that new technologies align with democratic governance, civil liberties, and public expectations.