Abstract

Excerpted From: Landon Brickey, Police And AI: When Abundantly Helpful Becomes Intrinsically Harmful, 91 Brook. L. Rev. 627 (Winter, 2026) (279 Footnotes) (Full Document)

Anyone who has watched an American crime drama has heard these immortal words countless times. In 1966, the landmark case Miranda v. Arizona extended the Fifth Amendment’s protection against self-incrimination and, thus, “the right to remain silent” to custodial interrogations. No longer were police officers allowed to interrogate a suspect in custody without informing them of their constitutional rights. But those rights stopped short of forbidding police officers from lying to suspects.

In 2018, Thomas Perez called the Fontana Police Department to report that his father, with whom he lived, was missing. The officer who received the call noted that Perez seemed distracted and uninterested, and when the police arrived at Perez’s home, they requested that Perez come to the police station for questioning without informing him that he was considered a suspect. While being questioned, the police obtained a search warrant and searched Perez’s home, uncovering microscopic bloodstains throughout the residence and detecting the smell of a dead body.

The evidence of the blood and smell prompted an approximately seventeen-hour interrogation that left Perez broken. During this time, and in addition to multiple interrogations, the police drove Perez around the city, asking him where he had dumped his father’s body. The police later told Perez that they had his father’s body in the morgue, and that the corpse was riddled with stab wounds. During one interrogation session, the police officers berated Perez, telling him he had killed his father, explaining how he did so, and asserting that Perez just could not remember the murder. Perez continued to claim his innocence. Then, the police brought in Perez’s dog and informed Perez he should say goodbye. The police officers went on to claim that the dog had to be euthanized because of the trauma it endured watching Perez kill his father. Over time, Perez ripped apart his shirt, hit himself, and tore out his hair. Shortly thereafter, Perez confessed to killing his father. When the police left the room, Perez unsuccessfully attempted to hang himself with his shoelace. The police put Perez on a seventy-two-hour psychiatric hold and sent him to a hospital.

However, there was one problem with the police’s case, or lack thereof: Perez’s father was alive. Shortly before Perez was placed on the seventy-two-hour hold, the police spoke with Perez’s sister, who told them that their father was visiting her upstate. The bloodstains found around Perez’s home were remnants of pinpricks from his father’s diabetes monitor. Perez was not informed that his father was alive while he finished his psychiatric hold and would not find out until almost three days later. Perez was never charged with a crime, and although he won nearly $900, from a civil lawsuit against the city, he may never psychologically recover, and the issue of police deception in interrogations remains. Perez’s case illustrates the extent to which some police will go to elicit a confession, and such police conduct is often imperfectly analyzed by the courts.

In determining if a defendant’s confession is allowed into evidence under the Due Process Clause of the Fifth and Fourteenth Amendments, courts apply the “totality of the circumstances” test, weighing several factors to deduce whether the confession was made voluntarily. One such factor in the totality of the circumstances test is whether a police officer has lied to a suspect. In determining an individual’s rights, the Miranda court stopped short of forbidding police officers from lying, tricking, or deceiving a suspect. Thereafter, lower courts have had to determine whether any use of deception overcame a suspect’s will. Merely three years after Miranda, in Frazier v. Cupp, the Supreme Court held that a suspect’s confession was voluntary under the Due Process Clause, even though a police officer induced the confession by falsely telling the suspect that his co-conspirator had confessed and implicated him in the crime. Since Frazier, police deception techniques, including outright lying, have been widely regarded as constitutionally permissible.

Although claims and exonerations from false confessions have been on the rise, courts allow such deception for several reasons, including efficiency, getting criminals to confess, and because most courts believe an innocent suspect would not confess to something they did not do. Courts typically allow these confessions into evidence unless other factors suggest the admission was involuntary. Consequently, many federal and state courts do not consider police deception to be the ultimate determinant in a “voluntariness” question. However, a few state courts have found that some deception techniques go too far. For example, the Second District Court of Appeal of Florida found that police officers creating fake laboratory results to elicit a confession in a statutory rape case was an unconstitutional violation of the defendant’s due process rights. Yet, in a Virginia case, a court found it was constitutional to use a defendant’s confession even when police showed him forged DNA and fingerprint tests linking him to the crime. The range of holdings and reasonings across jurisdictions is striking.

With the evolution of technology, including the accelerated growth and use of artificial intelligence (AI), the potential for false evidence to be used to elicit confessions is alarming. Recent legal and scholarly literature has grappled with the police’s use of AI in predictive policing, crime prevention, bail hearings, generating crime reports, and the criminal justice system generally. As AI technology continues to proliferate and as its capabilities increase, there may be excitement, but greater dangers are at play. For instance, proponents laud AI’s capacity to assist police in predicting when and where crimes might happen, in addition to its administrative efficiency. Conversely, many organizations have highlighted the discriminatory applications of predictive policing, such as built-in bias and over-policing of communities of color, as well as constitutional privacy concerns. The possible complications and benefits associated with AI affect nearly all aspects of policing, including the interrogation room. Although there are currently no public reports of police utilizing AI for deceptive interrogation techniques, the rapid development and use of AI in policing makes it clear that state legislatures should, at the very least, consider preventative measures for problems that AI deception could create.

This Note argues that state legislatures ought to forbid police from using AI to create false evidence to elicit confessions in police interrogations before they have the chance to. If police are allowed to use AI to manufacture evidence, it will be detrimental to society and criminal defendants, leading to an increase in false confessions, tainted evidence, and mistrust in the criminal justice system. Part I of this Note outlines Fifth Amendment jurisprudence surrounding Miranda, the “voluntariness” standard, and Miranda’s impact in the last sixty years. Part II examines the constitutionality of and reasoning behind various police deception tactics in federal and state case law through the standards and tests laid out in Part I. Part III discusses contemporary literature on AI in the criminal justice system and expands on the potential consequences of utilizing AI to aid in eliciting confessions through deceptive tactics. Part IV analyzes how courts and legislatures have begun grappling with the use of AI in the criminal justice system and identifies the approach that is best equipped to mitigate the adverse effects of leveraging AI as a police deception tactic.

[ . . . ]

The implications of artificial intelligence for the world are enticing and nerve-wracking. While AI has the potential to improve lives and support law enforcement, it also poses constitutional risks. New technology can evoke a range of emotions and fears, but society would not be where it is today if it were willing to take steps into the unknown. However, that does not mean we need to let AI run rampant, especially in the criminal justice system.

Thoughtful legislation and regulations can enable AI to make police work more efficient and communities safer without infringing on constitutional rights. Although that work will take time, there is no reason to wait for law enforcement to strip away individual rights before they are put back into place. There must be a push toward common-sense AI regulation that, in part, prohibits the use of AI in police interrogations until more can be done to prevent bias and false confessions.

J.D. Candidate, Brooklyn Law School, 2026; B.J., T