How AI Can Lead to False Arrests and Wrongful Convictions

On October 20, 2025, Taki Allen, a 17-year-old Baltimore student, was handcuffed at gunpoint after an AI-enhanced surveillance camera falsely identified a Doritos bag in his pocket as a gun. Police arrived within moments, searched him, and found only chips, turning a routine evening into a traumatic confrontation. The incident highlights how AI misidentification can lead to severe consequences when human judgment fails to account for uncertainty. Separately, Angela Lipps, a Tennessee grandmother, was arrested at gunpoint while babysitting her grandchildren on December 24, 2025. Facial recognition software had incorrectly linked her to fraud crimes in North Dakota, a state she had never visited. She spent five months in jail before her release, demonstrating how flawed AI tools can disrupt lives without regard for factual accuracy. Researchers studying AI in policing note that these systems operate on probabilities, not certainties. When AI tools predict crime risks or identify suspects, the statistical outputs are often treated as definitive evidence. For example, predictive policing algorithms analyze historical crime data to score neighborhoods, routing officers to high-risk areas. However, the uncertainty inherent in these predictions is frequently overlooked, turning statistical likelihoods into operational decisions with real-world consequences. Generative AI models like ChatGPT similarly generate probable responses rather than verified facts. While they may produce correct answers, such as naming Thomas Edison as the inventor of the light bulb, they lack context—ignoring contributions like Joseph Swan’s parallel work. This probabilistic nature poses risks in law enforcement, where AI-generated predictions can be misinterpreted as irrefutable truth. The issue extends beyond individual cases. AI policing tools are deployed in dozens of U.S. cities, though no public registry tracks their full use. These systems often rely on biased historical crime data, reinforcing cycles of over-policing in marginalized communities. The shift from probabilistic prediction to operational certainty occurs rapidly, eroding accountability and increasing the likelihood of wrongful arrests. Experts emphasize that treating AI outputs as facts rather than educated guesses undermines justice. Without transparency and rigorous oversight, flawed AI systems will continue to cause harm, disproportionately affecting vulnerable individuals.