Multiple outlets discuss how AI systems used in policing can contribute to false arrests and wrongful convictions when their probabilistic outputs are treated as definitive evidence. The Conversation highlights that risk increases when law enforcement assumes AI returns “certainties,” rather than likelihoods, which can distort decision-making and strengthen inaccurate conclusions. Salon similarly emphasizes that common AI tools—such as facial recognition systems—produce probabilities rather than facts. In both accounts, the core problem is interpretation: if probabilities are presented, understood, or relied on as if they were conclusive, they can influence investigators, prosecutors, and courts toward outcomes that do not reflect the underlying uncertainty of the model. The reporting focuses on the mismatch between how AI results are generated and how they may be used in legal settings, suggesting that misunderstanding AI outputs can have serious downstream consequences. Overall, the sources converge on the idea that careful handling of AI uncertainty is necessary to avoid wrongful outcomes.