Police use of artificial intelligence raises constitutional challenges under the Fourth Amendment (warrantless surveillance), the First Amendment (chilled speech and association), the Fourteenth Amendment (biased outcomes in protected classes), and the Fifth and Sixth Amendments (the ability to challenge algorithmic evidence in court). No single federal law resolves these questions. Instead, protections come from a scattered mix of Supreme Court decisions, state statutes, and local ordinances, and what shields you in one city may not exist a few miles away.
The tools driving these questions are already in wide use. Automated license plate readers photograph every plate that passes a camera and log time, location, and direction of travel. Facial recognition matches images against databases of mugshots and driver’s license photos. Predictive systems forecast where crimes will occur or score individuals for risk of reoffending. Acoustic sensors like ShotSpotter (now SoundThinking) locate gunfire and generate reports that prosecutors introduce as evidence. Each of these tools sits in a different constitutional pressure point.
Fourth Amendment: When Does AI Surveillance Require a Warrant
The Fourth Amendment protects “the right of the people to be secure in their persons, houses, papers, and effects, against unreasonable searches and seizures.”1Constitution Annotated. U.S. Constitution – Fourth Amendment The core question for AI surveillance is whether deploying these tools against you counts as a “search” at all.
Under Katz v. United States, a search occurs when the government violates a privacy expectation society is prepared to recognize as reasonable. The Court reinforced that framework in Kyllo v. United States, striking down the warrantless use of a thermal imaging device pointed at a home and warning that permitting technology-assisted surveillance from public vantage points “would leave the homeowner at the mercy of advancing technology.”2Constitution Annotated. Katz and Reasonable Expectation of Privacy Test
Cutting the other way is the third-party doctrine, which long held that people have no reasonable expectation of privacy in information they share with someone else. That doctrine gave law enforcement broad, warrantless access to bank records, phone logs, and similar data for decades.
The Supreme Court pulled back in Carpenter v. United States, holding that acquiring long-term cell-site location records is a Fourth Amendment search requiring a warrant. The Court declined to extend the third-party doctrine to cell-site location information, finding that it “implicates even greater privacy concerns than GPS tracking” and that people do not meaningfully volunteer their location data to cell carriers simply by carrying a phone.3Supreme Court of the United States. Carpenter v. United States Syllabus
Carpenter opened the door to what scholars call the mosaic theory. A single license plate reader photograph of your car at a gas station is trivial. Millions of such photographs, cross-referenced across time and location, can reveal where you live, where you work, where you worship, who you visit, and whether you attended a political rally. Lower courts are now applying this reasoning to challenge warrantless access to ALPR databases, GPS records, and other forms of continuous AI-enabled surveillance. The constitutional question is no longer whether any single data point is private, but whether the aggregate portrait requires a warrant.
First Amendment: Surveillance and the Chilling Effect
AI surveillance also raises First Amendment problems. When people know facial recognition or license plate readers may identify them, they are less likely to attend protests, join controversial organizations, or express unpopular views. In NAACP v. Alabama, the Supreme Court held that compelled disclosure of group membership can violate freedom of association, especially for organizations that espouse dissenting views. In Americans for Prosperity Foundation v. Bonta, the Court held that compelled disclosure of affiliations with advocacy groups must be narrowly tailored, and that the burden on associational rights is cognizable even without direct government retaliation.
The doctrine is uneven. In Laird v. Tatum, the Supreme Court held that a subjective sense of being watched was not enough to challenge a military surveillance program. Lower courts have found actionable chilling effects where surveillance data was disclosed publicly or lacked adequate safeguards. At a protest, facial recognition can identify attendees and ALPR databases can log who drove there. The First Amendment harm is not just what the government does with that information but the deterrent effect of knowing it exists.
Equal Protection: Algorithmic Bias and the Intent Problem
The Fourteenth Amendment guarantees that no state shall “deny to any person within its jurisdiction the equal protection of the laws.”4Constitution Annotated. Amdt14.S1.3 Due Process Generally AI systems strain this guarantee in ways existing doctrine handles poorly.
The evidence of disparate performance is concrete. A 2019 National Institute of Standards and Technology study tested 189 facial recognition algorithms from 99 developers and found that false positive rates varied by demographic group by factors of 10 to 100. The highest false positive rates appeared in West African, East African, and East Asian faces; the lowest appeared in Eastern European faces. On domestic law enforcement mugshot images, false positives were highest for American Indian individuals, with elevated rates for African American and Asian populations. Several algorithms developed in China showed reversed patterns, performing better on East Asian faces, which suggests the disparities stem largely from training data composition rather than anything inherent to the technology.5NIST. Face Recognition Vendor Test (FRVT), Part 3: Demographic Effects
Predictive policing produces a related pattern through feedback loops. If a neighborhood was heavily policed for years, its historical arrest data will reflect that, and an algorithm trained on that data will flag the neighborhood as high-risk. More officers show up, make more arrests, and the cycle reinforces itself.
The legal problem is doctrinal. Equal protection claims based on race trigger strict scrutiny, but only if the plaintiff can establish discriminatory purpose, not just discriminatory effect. An algorithm that produces racially disparate outcomes is not necessarily unconstitutional if no one designed it with discriminatory intent. That intent requirement is a poor fit for systems where biased outcomes emerge from data patterns rather than deliberate choices, and courts have not yet resolved the tension.6Harvard Law Review. Beyond Intent: Establishing Discriminatory Purpose in Algorithmic Risk Assessment
Due Process: Challenging AI Evidence in Court
When AI-generated evidence supports a prosecution, the Fifth and Sixth Amendments raise pointed questions about what the defense can actually challenge.
The Confrontation Clause Gap
The Sixth Amendment guarantees criminal defendants the right to confront the witnesses against them. Courts have almost universally held that this right does not extend to machine-generated evidence, because machines are not “witnesses” in the constitutional sense. In Commonwealth v. Weeden, a Pennsylvania court upheld the admission of a ShotSpotter report identifying the time, location, and number of gunshots, reasoning that no individual authored the automatically generated report and no confrontation right applied. The Seventh Circuit has said “data are not ‘statements’ in any useful sense,” and the Eleventh Circuit has held that “the Confrontation Clause is concerned with human witnesses.”
This leaves a practical gap. When a human analyst testifies, defense attorneys can probe qualifications, methods, and bias. When a machine generates the evidence, there is often no one to cross-examine about error rates, training data quality, or calibration. Justice Sotomayor flagged this concern in her concurrence in Bullcoming v. New Mexico, noting that purely machine-generated results raise different issues than results involving human analysis. Lower courts have occasionally pushed back. A California appellate court reversed a conviction partly because ShotSpotter evidence was admitted without any expert testimony about the system’s margin of error, training protocols, or mathematical models.
Risk Scores and Sentencing
The Wisconsin Supreme Court addressed algorithmic risk assessment in State v. Loomis, ruling that using the COMPAS tool at sentencing did not violate due process, but imposing significant restrictions. Judges cannot use risk scores to decide whether someone is incarcerated or to set sentence severity. Any presentence report incorporating COMPAS must include written warnings about the tool’s limitations, including its proprietary nature, its reliance on group-level data, and studies questioning whether it disproportionately classifies minority offenders as higher risk.
Brady and Proprietary Algorithms
Under Brady v. Maryland, prosecutors must disclose evidence favorable to the defendant when it is material to guilt or punishment.7Justia. Brady v. Maryland, 373 U.S. 83 (1963) Applied to AI, this raises whether prosecutors must disclose source code, training data, error rates, and known limitations of any algorithm that contributed to the case. If facial recognition identified you and that system has elevated false positive rates for your demographic group, that information is arguably material.
The tension is real. Many AI tools are proprietary, and developers resist disclosing source code as a trade secret. The Wisconsin Supreme Court acknowledged that COMPAS’s proprietary nature prevents anyone from knowing exactly how scores are calculated, but upheld its use with warnings rather than mandating disclosure. A New Jersey appellate court took a different route in New Jersey v. Arteaga, holding that defendants must be notified when facial recognition was used in the investigation that led to charges, so they can review potentially exculpatory information.
What Federal Law Currently Requires
The federal regulatory picture is unsettled. In October 2023, the Biden administration issued Executive Order 14110, which imposed safety and transparency requirements on federal agencies using AI. In January 2025, the Trump administration revoked that order, describing its requirements as “barriers to American AI innovation” and directing agencies to review policies issued under it for possible suspension or rescission.8The White House. Removing Barriers to American Leadership in Artificial Intelligence
Some risk management requirements survive. An OMB memorandum requires federal agencies to complete impact assessments for high-impact AI use cases, with agencies required to update procurement policies for AI systems by early 2026.9The White House. Increasing Public Trust in Artificial Intelligence Through Unbiased AI Principles The FBI classifies several of its AI systems as high-impact, including biometric matching across large datasets, automatic vehicle identification, and language translation tools. As of early 2026, none of the FBI’s nine high-impact AI use cases had completed the required impact assessments, and the Department of Justice acknowledged that its AI inventory remained incomplete. Congress has introduced targeted bills, but no comprehensive federal AI policing statute has been enacted.
State and Local Restrictions
With federal action stalled, state and local governments have imposed most of the binding rules. More than a dozen states now regulate law enforcement facial recognition in some form. Common restrictions include:
- Requiring a warrant or probable cause before using the technology
- Limiting its use to investigations of serious crimes
- Requiring that defendants be notified when facial recognition contributed to their case
- Prohibiting arrests based solely on a facial recognition match
- Setting accuracy testing standards
- Banning the use of facial recognition with body-worn cameras
Cities including Oakland, Pittsburgh, New Orleans, and Santa Cruz have enacted outright bans on predictive policing. Others have adopted surveillance oversight ordinances that require community approval before police acquire new AI tools. The Detroit Police Department agreed through a legal settlement to detailed restrictions on facial recognition use, including mandatory training on the technology’s risks, a requirement to corroborate any match with independent evidence before placing someone in a photo lineup, and an obligation to inform defendants when facial recognition was used.
The result is a patchwork. A tool that is banned in one city may be standard equipment in the next jurisdiction over, with no federal floor establishing minimum protections.
If You Are Affected by Police AI
If you are charged with a crime and suspect AI contributed to the investigation, several points of leverage exist depending on where you are. Ask whether facial recognition or another AI tool was used to identify you; in New Jersey and jurisdictions like Detroit, that notification is required. Ask whether any match was independently corroborated before an arrest or lineup. If a risk assessment score influenced pretrial detention or sentencing, Loomis requires warnings about the tool’s limitations, and judges cannot use the score to determine whether you are incarcerated or how severe the sentence will be. If a machine-generated report is being introduced as evidence, courts are divided on what expert testimony must accompany it, and some have reversed convictions where none was offered.
Even the strongest current view of these tools treats them as advisory. An officer should not arrest someone solely because an algorithm flagged them, and a judge should not set bail solely based on a risk score. Whether agencies follow that principle varies, and enforcement mechanisms remain thin. The constitutional questions raised by AI in policing are still being litigated, and the answers depend heavily on the jurisdiction, the tool, and the amendment invoked.