Leaders of a Beautiful Struggle v. Baltimore Police Department

Executive Summary

Leaders of a Beautiful Struggle v. Baltimore Police Department is one of the most important federal appellate decisions applying the Fourth Amendment to persistent surveillance technology. Sitting en banc, the Fourth Circuit held that Baltimore’s Aerial Investigation Research program violated the Fourth Amendment because the system permitted police to reconstruct and analyze the movements of individuals across the city over time. The court treated the program not as a collection of isolated photographs taken from lawful airspace, but as a retrospective database capable of revealing the “whole” of a person’s movements when combined with investigative analysis and other surveillance systems.

The case arose from Baltimore’s decision to conduct a six-month pilot program with Persistent Surveillance Systems, a private contractor. Multiple aircraft flew over Baltimore carrying wide-area imaging cameras. The system captured roughly thirty-two square miles in each image, at approximately one image per second, and provided an estimated twelve hours of coverage over about ninety percent of the city on operating days, weather permitting. The contractual program limited collection to daylight hours and limited resolution to approximately one pixel per person or vehicle. Individual people therefore appeared as dots or blobs rather than identifiable faces.

Those limitations were central to the defense of the program. Baltimore argued that the imagery was much less intrusive than ordinary cameras because it could not identify faces, read license plates, see inside buildings, or continuously follow a person through the night. Analysts generally worked retrospectively rather than in real time. The district court accepted much of that reasoning and denied a preliminary injunction. A divided Fourth Circuit panel affirmed.

The en banc majority viewed the technology differently. Chief Judge Gregory emphasized that the relevant Fourth Amendment question was not what a single aerial image showed. The plaintiffs challenged the creation and use of a retrospective record of movements across the city. The program’s analytical value came from linking successive images, following movement from crime scenes to other locations, and integrating the aerial record with Baltimore Police Department systems such as CitiWatch cameras, license plate readers, ShotSpotter, dispatch information, and other investigative data. Once police linked a “dot” to a person or vehicle, the low resolution did little to prevent reconstruction of where that person traveled.

The majority relied heavily on Carpenter v. United States, United States v. Jones, and the D.C. Circuit’s pre-Jones decision in United States v. Maynard. Those cases recognized that aggregated location information can reveal facts that no single observation discloses. A sequence of movements may expose home, work, relationships, political affiliations, religious practices, medical treatment, and other intimate details. The en banc court concluded that the AIR program’s persistent, retrospective character placed it within that line of cases.

In the majority’s view, the program’s daylight limitation did not solve the constitutional problem. Most movement occurs during the day, and people frequently begin and end their days at home. Analysts could sometimes re-identify a target across days by using contextual clues and other police systems. Nor did the program’s one-pixel resolution eliminate privacy concerns. The court reasoned that identity can often be inferred from patterns: a track ending at a residence, a vehicle later appearing on an ALPR system, or a ground camera capturing the same person or car.

The court also rejected the idea that ordinary aerial-surveillance precedents controlled. Traditional cases involving aircraft or helicopters generally addressed discrete visual observation of particular properties or targets. The AIR program instead created a searchable, retrospective record covering much of a city. The majority considered that difference constitutionally meaningful. In its words, the aerial nature of the technology was “incidental” to the plaintiffs’ claim; the controlling concern was privacy in physical location and movement.

The litigation included a major procedural complication. While the appeal was pending, the six-month pilot ended, the planes stopped flying, Baltimore decided not to renew the program, and much of the collected imagery was deleted. The defendants argued that the case was therefore moot. The en banc court disagreed because Baltimore and the contractor retained imagery, investigation reports, and related data connected to specific cases, and the plaintiffs had sought to enjoin both collection and access. Because the police department continued to possess and use retained material, the controversy remained live.

The majority reversed the district court’s denial of preliminary relief. It held that accessing AIR data was a search and that the program’s warrantless operation violated the Fourth Amendment. The court further found the remaining preliminary-injunction factors satisfied, emphasizing that the loss of constitutional rights constitutes irreparable injury and that government has no legitimate interest in maintaining a program likely to violate the Constitution.

The decision produced extensive separate opinions. Chief Judge Gregory wrote separately to respond to concerns about crime, policing, and community choice. Judge Wynn wrote a concurrence emphasizing race, policing, and the unequal burdens surveillance can impose on heavily policed communities. Judge Wilkinson authored a lengthy dissent joined in varying parts by several judges, arguing that the case had become moot, that the majority overstated the program’s capabilities, that the district court deserved deference, and that the majority improperly extended Carpenter and the mosaic theory. Judge Niemeyer separately dissented, and Judge Diaz wrote that he viewed the underlying Fourth Amendment question as close but would have dismissed the case as moot.

For public-safety agencies, the case is significant well beyond aircraft. The court’s reasoning focuses on surveillance architecture: persistence, retrospective access, geographic scale, data integration, and the ability to infer identity or patterns from aggregated observations. Those same features increasingly appear in automated license plate reader networks, real-time crime centers, drone programs, fixed camera networks, connected-vehicle systems, and AI-powered analytics. A technology may collect only information exposed to public view at each moment yet still raise constitutional risk if the system creates a detailed, durable, searchable record of movement at a scale that ordinary observation could not practically achieve.

Five Key Takeaways

Why Leaders of a Beautiful Struggle Matters

It moves the constitutional inquiry from the sensor to the system

The decision refuses to analyze persistent surveillance solely by asking what one camera frame reveals. A single AIR image showed only blurred dots. The constitutional problem arose from the system’s ability to connect those dots across time and combine them with other sources. For agencies procuring modern analytics, this distinction is critical: constitutional risk can arise from the system-level capability even when each sensor seems modest in isolation.

It is one of the clearest appellate applications of the mosaic concept

The “mosaic theory” describes the idea that a series of individually innocuous observations can collectively reveal information entitled to constitutional protection. The Supreme Court has not adopted that label as a universal doctrinal test, but the reasoning of Jones and Carpenter gave aggregation a central role in modern location privacy. The Fourth Circuit applied that logic directly to a citywide surveillance program.

It treats retrospective access as constitutionally important

The program’s power was not principally real-time viewing. Its value was the ability to go back after a serious crime and reconstruct movement. That retrospective quality resembles historical CSLI and modern data-fusion platforms. It changes law enforcement from “watching what happens” to “searching what already happened.”

It anticipates integrated surveillance

The AIR contract contemplated integration with ground cameras, ALPR, gunshot detection, and dispatch systems. That architecture resembles modern real-time crime centers. The opinion therefore provides a useful framework for thinking about the constitutional effect of integrated systems rather than isolated technologies.

It is especially important after Chatrie

The Supreme Court’s 2026 decision in Chatrie v. United States reinforces the Fourth Circuit’s emphasis on retrospective movement data. Chatrie held that even two hours of precise Location History can constitute a search and rejected a categorical short-duration exception. While the technologies differ, both decisions caution against assuming that public movement or limited duration automatically removes location analytics from Fourth Amendment scrutiny.

Technology Background: Baltimore’s AIR Program

Baltimore’s Aerial Investigation Research program used aircraft equipped with Persistent Surveillance Systems’ Hawkeye Wide Area Imaging System. Multiple planes flew separate orbits above Baltimore. According to the en banc opinion, the cameras captured roughly thirty-two square miles per image per second. The aircraft flew at least forty hours each week and could provide approximately twelve hours of coverage over roughly ninety percent of the city on an operating day, weather permitting.

The contract imposed two major limits. Collection occurred during daylight hours, and resolution was restricted to approximately one pixel per person or vehicle. Those restrictions were policy choices rather than inherent technical limitations. A person or car could be seen as a small dot or blob, but the aerial image generally did not reveal face, clothing detail, or a readable license plate.

The program targeted specified serious crimes: homicide and attempted murder, shootings involving injury, armed robbery, and carjacking. Between fifteen and twenty-five contractor analysts worked in shifts. Rather than continuously watching for crimes in real time, analysts ordinarily responded after Baltimore officers requested support in a qualifying investigation.

PSS aimed to provide an initial briefing within eighteen hours and a fuller investigation briefing report within seventy-two hours. Analysts could trace vehicles and people from the crime scene, identify locations visited, observe driving patterns, and follow connections to other people and locations. Limited real-time support was technically possible in exigent circumstances with a written request from the police commissioner.

Low resolution did not mean low analytical power

The majority treated the distinction between image resolution and tracking capability as fundamental. The imagery did not need to identify a person’s face if police could infer identity by tracking a dot to a house, vehicle, camera, or other identifying context. The program’s objective was precisely to turn anonymous movement into investigative identity.

City-scale imagery altered ordinary surveillance assumptions

Traditional fixed cameras observe one intersection, building, or block. AIR captured a broad geographic field in each frame. That meant analysts could move backward and forward through stored imagery and follow a path across neighborhoods without needing a camera prepositioned at every turn.

Facts and Program Development

Baltimore’s relationship with Persistent Surveillance Systems predated the 2020 AIR pilot. In 2016, the public learned that Baltimore police had been using aerial surveillance technology in partnership with PSS. Public controversy followed, and that earlier program was discontinued.

In December 2019, Police Commissioner Michael Harrison announced a renewed partnership. This time the department proposed a six-month Aerial Investigation Research pilot. Baltimore planned public meetings before implementation. The first occurred in March 2020; additional meetings were disrupted by the COVID-19 pandemic, and the department streamed presentations online.

On April 1, 2020, Baltimore’s Board of Estimates approved the professional services agreement with PSS. The requested funding—about $3.69 million—came from private philanthropic support rather than the city budget.

The plaintiffs were community advocates whose work regularly required them to travel through Baltimore and be present near scenes of violence. Leaders of a Beautiful Struggle, Erricka Bridgeford, and Kevin James filed suit on April 9, 2020. They challenged the program under the Fourth Amendment and sought to prevent both collection and access to the surveillance data.

The district court denied preliminary relief on April 24, 2020. The AIR planes began operations approximately one week later. The plaintiffs appealed immediately.

Data Retention, Investigation Reports, and System Integration

Forty-five-day imagery retention

The professional services agreement stated that raw AIR imagery would be retained for forty-five days. Reports and related images could be maintained longer for legal proceedings and applicable limitations periods. The majority later noted that actual retention practices during the pilot were more complicated than the simplified forty-five-day description.

After the pilot, Baltimore initially held 1,916.6 hours of coverage comprising more than 6.6 million images. In January 2021, Baltimore and PSS deleted most of the imagery but retained material linked to investigations. The retained subset included hundreds of hours and hundreds of thousands of cropped images, as well as investigation briefings and ground-based video uploaded to the department’s evidence system.

Integration with other police systems

The contract contemplated integration with Baltimore’s dispatch system, CitiWatch camera network, ShotSpotter, and license plate readers. This mattered to the majority because integrated systems make re-identification easier. If an aerial track reaches a road monitored by ALPR or a ground camera, another sensor can supply identifying information that the aerial image lacks.

The system therefore operated as more than a standalone aerial camera. It was designed to function as a layer within an investigative data environment. The majority saw this integration as evidence that the practical capability of AIR had to be assessed in context.

Retention and reuse created a continuing controversy

The retention of investigation-related material became important to Article III mootness. Even though planes stopped flying and Baltimore ended the pilot, the police department retained data relevant to approximately two hundred investigations and continued to possess reports and associated materials. Because the plaintiffs had asked to enjoin access as well as collection, the majority concluded that meaningful relief remained possible.

Procedural History

Mootness: Why the Case Remained Live After the Pilot Ended

Baltimore argued that the case became moot because the planes had stopped flying, the city terminated the contract, most imagery had been deleted, and there would be no new data collection under the AIR pilot. If no live controversy remained, Article III would have required dismissal without reaching the constitutional question.

The majority rejected that position. Plaintiffs had asked to prohibit not only future collection but also access to existing AIR data. Baltimore retained investigation-related imagery and reports. The police department therefore still possessed material that plaintiffs claimed had been generated and could be accessed in violation of the Fourth Amendment.

The court concluded that an injunction barring access to retained data could provide concrete relief. That was enough to keep the case live. The decision is a reminder that ending a surveillance program does not necessarily moot litigation when agencies retain data collected during the program and continue to use those records in investigations or prosecutions.

Operational implication

Sunsetting technology should include a legal plan for retained data. Agencies should identify what will be deleted, what must be preserved for evidentiary obligations, what remains accessible, who may use it, and whether ongoing use creates continuing constitutional or statutory exposure.

Holding

The en banc Fourth Circuit held that accessing AIR data constituted a Fourth Amendment search because the program enabled law enforcement to deduce from the whole of individuals’ movements. The court further held that the program’s warrantless operation violated the Fourth Amendment.

The majority did not rest on physical trespass. Nor did it say that every aerial photograph constitutes a search. Instead, it applied the reasonable-expectation-of-privacy reasoning developed in Carpenter, Jones, and Maynard to a technology capable of creating a retrospective movement record.

The court reversed the denial of the preliminary injunction and remanded. The result effectively prevented Baltimore from continuing to access retained AIR materials without satisfying Fourth Amendment requirements.

The En Banc Majority’s Fourth Amendment Analysis

The court defined the challenged search at the level of access to the database

The majority emphasized that plaintiffs were not challenging a single photograph or the mere fact that an airplane captured a person outdoors. Their claim targeted the creation and use of a retrospective movement database. The court therefore analyzed what accessing that database could reveal.

Traditional aerial surveillance cases were not dispositive

Baltimore relied on cases permitting visual observation from aircraft. The majority distinguished those cases because they involved discrete observations of particular places or targets. AIR was different in scale and persistence. It captured a broad portion of the city repeatedly and stored the record for later investigative reconstruction.

The majority stated that the aerial nature of the program was incidental to the core privacy claim. The controlling precedents concerned privacy in physical location and movements. That formulation is particularly important for future technologies because it prevents agencies from assuming that a sensor-specific line of precedent resolves a system-level aggregation question.

Daylight-only collection remained persistent enough

The defense emphasized that planes flew only during daylight and that tracks often lasted only a few hours. The majority rejected the idea that those limitations made AIR equivalent to ordinary short-term surveillance. Most people engage in most daily movement during daylight. People often begin or end at home, and recurring patterns can facilitate re-identification across days.

Identity could be inferred from context

One pixel per person did not prevent tracking. The court reasoned that analysts could follow movements, observe destinations, and combine the aerial record with ground cameras, ALPR, public information, and police databases. A dot that could not be visually recognized might still become identifiable through its path.

Aggregation revealed more than isolated observations

The majority invoked the insight that a sequence of movements can disclose much more than any individual trip. Home, workplace, religious practice, medical care, personal relationships, and political association can emerge from patterns. This was the core “whole of movements” reasoning drawn from Maynard, Jones, and Carpenter.

Practical surveillance capacity matters

The opinion also relied on the fact that wide-area aerial imagery surpassed ordinary expectations about what police can feasibly observe. Without the system, reconstructing citywide movements retrospectively would require extraordinary resources and luck. AIR made that capability systematic.

The Mosaic Theory and the “Whole of Movements”

The mosaic concept rests on a simple proposition: constitutional significance can arise from aggregation. A single observation that a person drove down a public street ordinarily reveals little. Repeated observations that identify where the person sleeps, works, worships, receives medical treatment, socializes, and meets others can reveal an intimate portrait.

The D.C. Circuit articulated this reasoning in United States v. Maynard, which preceded the Supreme Court’s GPS decision in Jones. Several Justices in Jones then suggested that prolonged location monitoring could violate reasonable expectations of privacy even without physical trespass. Carpenter later relied on the revealing nature of aggregated historical location information in holding that government acquisition of historical CSLI is a search.

Leaders of a Beautiful Struggle is one of the strongest federal appellate applications of that line of thought to visual surveillance. The majority did not label every series of camera observations a search. It focused on a system designed to create a citywide retrospective movement record and integrate it with other identifying information.

Why “public” does not end the analysis

Movements in public can be observed. But the majority treated the practical difference between occasional human observation and comprehensive retrospective reconstruction as constitutionally significant. The Fourth Amendment inquiry therefore did not end with the fact that AIR saw only outdoor movement.

Why integration matters

Aggregation can occur within one dataset or across datasets. AIR’s power increased when linked with ALPR, ground cameras, dispatch records, and other police systems. Modern agencies should expect courts to look at how systems actually work together rather than assessing each component in isolation.

Relationship to Carpenter, Jones, Maynard, and Knotts

The majority’s use of Carpenter

The Fourth Circuit reasoned that Carpenter was not merely about cell phones. It was about government access to records capable of reconstructing physical location and movement in a way that exceeds ordinary surveillance expectations. That functional reading allowed the majority to apply Carpenter to aerial imagery.

The dissents’ response

The dissents argued that the majority stretched Carpenter too far. AIR was less continuous, less precise in identity, limited to daylight, and dependent on analyst labor. In their view, those differences kept the program closer to permissible visual surveillance than to the comprehensive CSLI archive in Carpenter.

Concurring Opinions

Chief Judge Gregory’s separate concurrence

Chief Judge Gregory wrote separately to answer the broader policy arguments raised by the dissents. He rejected the framing that Baltimore residents had to choose between safety and constitutional privacy. The concurrence discussed the perspectives of community advocates and emphasized that constitutional rights do not disappear because a city faces serious violent crime.

The concurrence is not necessary to the Fourth Amendment holding, but it illustrates the context in which surveillance programs operate. Technology policy can implicate public safety, community trust, civil liberties, race, and governance simultaneously. Courts decide constitutional questions, not which constituency speaks for an entire community.

Judge Wynn’s concurrence

Judge Wynn wrote separately to address the history and distribution of policing burdens in Baltimore. His concurrence emphasized that surveillance does not operate in a social vacuum and can fall disproportionately on neighborhoods already subject to intensive police activity.

For agencies, the concurrence underscores a governance point distinct from the constitutional minimum: even a legally defensible technology may create legitimacy problems if deployment, retention, or analytics disproportionately burden particular communities without clear controls and transparency.

Dissenting Opinions

Judge Wilkinson

Judge Wilkinson authored the principal dissent, joined in varying parts by several judges. He would have dismissed the case as moot and criticized the majority for reaching the merits after Baltimore ended the program. He also argued that the majority improperly displaced the district court’s factual findings and overstated AIR’s surveillance capability.

On the Fourth Amendment, the dissent stressed the program’s limitations: daylight-only operation, low resolution, inability to see inside buildings, interruptions when a subject entered a structure, analyst labor, and the fact that tracks often lasted only a few hours. In the dissent’s view, those limits made AIR substantially less revealing than the long-term CSLI at issue in Carpenter.

Judge Wilkinson also emphasized federalism and local experimentation. Baltimore faced severe violent crime, and elected officials chose a pilot program with restrictions and evaluation mechanisms. The dissent cautioned against federal courts prematurely invalidating local public-safety experiments on a sparse preliminary-injunction record.

Judge Niemeyer

Judge Niemeyer separately dissented and criticized the majority’s treatment of the Fourth Amendment and the factual record. His position reflected a more traditional view that observation of movements exposed to public view generally does not create a reasonable expectation of privacy.

Judge Diaz

Judge Diaz described the constitutional question as close and acknowledged the strength of arguments on both sides. He nevertheless would have dismissed the appeal as moot after the city terminated the pilot and deleted most of the data.

Why the dissents matter operationally

The dissents identify the factual characteristics that may distinguish one persistent surveillance system from another. Coverage duration, resolution, ability to re-identify, analyst effort, retention, integration, geographic scope, and real-time capability can all affect constitutional analysis. Agencies should document those characteristics rather than relying on broad labels such as “camera,” “drone,” or “aerial surveillance.”

Important Limits of the Decision

Leaders of a Beautiful Struggle is broad but not limitless. It does not establish that every camera network, drone flight, helicopter observation, or ALPR query is a Fourth Amendment search.

It involved persistent, city-scale surveillance

The AIR system repeatedly captured large portions of Baltimore and preserved imagery for retrospective analysis. A brief targeted aerial observation presents materially different facts.

It involved retrospective tracking

The court repeatedly focused on the ability to go back in time and reconstruct movement after a crime. A live camera used to observe one scene does not necessarily create the same concern.

It involved integration

AIR could operate with ALPR, ground cameras, ShotSpotter, and dispatch systems. An isolated sensor with no historical database or cross-system linkage may present lower risk.

It involved warrantless operation

The court held the warrantless program unconstitutional. It did not decide what form of warrant, court order, or other judicial authorization would be sufficient for all uses of comparable technology.

It was a preliminary-injunction case

The procedural posture required the court to evaluate likelihood of success and equitable factors rather than resolve a criminal suppression motion after a full trial record. That context shaped parts of the disagreement among the judges.

Practical Guidance for Law Enforcement

Assess the program as a system

Before deployment, map every data source, integration, retention rule, search capability, and inference the platform can generate. Do not evaluate each sensor in isolation if the operational design is to combine them.

Distinguish live monitoring from retrospective reconstruction

Policies should separately define real-time observation, historical review, pattern-of-life analysis, and reverse searching. Each capability can raise different legal and policy issues.

Limit duration and geographic scope

Persistent surveillance should be tied to a defined mission, location, and period. Citywide or indefinite collection poses greater risk than targeted deployment.

Define when identification is permitted

Systems often begin with anonymous or pseudonymous movement. Policy should specify what evidentiary threshold permits an analyst to use ALPR, camera, subscriber, or other records to identify the person behind the track.

Control cross-system fusion

Integration can transform low-detail data into high-detail intelligence. Agencies should require legal review before linking systems and should document what new capability the integration creates.

Build retention around mission need

Long retention increases retrospective surveillance power. Set default deletion periods and define exceptions for evidentiary preservation, litigation holds, or active investigations.

Preserve auditability

Every historical query should identify who searched, what criteria were used, what data sources were accessed, what results were produced, and why the search was authorized.

Use heightened review for sensitive locations

Persistent surveillance around homes, clinics, religious institutions, political meetings, news organizations, schools, and other sensitive places should receive additional supervisory or legal review.

Do not equate low resolution with low privacy impact

An image can be visually anonymous yet analytically identifiable. Policies should evaluate inference and re-identification, not merely face resolution.

Persistent Surveillance Program Review Checklist

1. Mission: What specific crimes or public-safety problems may the system address?
2. Geographic scope: What area is covered, and why is that area necessary?
3. Temporal scope: How many hours per day and how many days or months will collection occur?
4. Resolution: What can a single sensor observation reveal?
5. Persistence: Can a person or vehicle be followed across time?
6. Retrospective capability: Can analysts reconstruct events after the fact?
7. Identification: Can anonymous tracks be tied to persons, vehicles, homes, or accounts?
8. Integration: Is the system connected to ALPR, cameras, CAD, gunshot detection, facial recognition, social media, or commercial data?
9. Retention: How long is raw data retained?
10. Derived records: How long are reports, tracks, alerts, and extracted clips retained?
11. Reuse: Can data retained for one case later be searched for another?
12. Legal process: What searches require a warrant, court order, subpoena, or supervisory approval?
13. Sensitive locations: Does the system reveal attendance at constitutionally sensitive places?
14. First Amendment activity: What rules prevent monitoring of lawful political, religious, or associational activity?
15. Minimization: How is information concerning uninvolved persons handled?
16. Audit logs: Are all historical searches and exports logged?
17. Vendor access: Can contractors search or reuse the data?
18. Model or analytics training: May vendor algorithms learn from agency data?
19. Community transparency: What information about the program is public?
20. Sunset and deletion: What happens if the pilot or contract ends?

AI and Data-Fusion Implications

The AIR program anticipated the architecture now emerging in AI-enabled policing. Its most important capability was not the aircraft alone. It was the ability to combine aerial imagery with other information to follow tracks, infer identity, and reconstruct events. Modern artificial intelligence can automate nearly every one of those steps. That makes Leaders of a Beautiful Struggle increasingly relevant even though the specific Baltimore AIR pilot no longer operates.

AI lowers the cost of mosaic surveillance

Persistent surveillance once required teams of analysts manually reviewing imagery, comparing tracks, consulting dispatch records, and locating ground-camera footage. Computer vision can now perform object detection and object tracking automatically. Entity-resolution systems can associate a track with a plate, vehicle, residence, account, prior incident, or known person. Generative systems can summarize the resulting pattern in seconds. As analytical cost falls, practical friction becomes a weaker constraint on surveillance scale.

That matters because the Fourth Circuit’s reasoning repeatedly focused on what technology makes practically possible. If a surveillance system allows an agency to reconstruct movements that previously would have required extraordinary human effort, courts may view the change in scale as constitutionally significant even if the underlying observations occurred in public.

Object detection and automated tracking

Computer-vision systems can detect people, vehicles, bicycles, and other objects across successive frames. A modern system may automatically assign an object identifier and maintain a track through an area. When coverage overlaps across cameras or sensors, software can attempt to continue the track across different fields of view.

Automated tracking changes the role of the analyst. Instead of deciding frame by frame where a person went, the analyst may begin with a preassembled movement path. Agencies should therefore treat object-tracking capability as part of the surveillance system’s privacy impact, even if the raw imagery itself remains low resolution.

Person re-identification

Person re-identification technologies attempt to determine whether the same person appears in different camera views even when facial recognition is unavailable or unsuitable. Systems may rely on clothing, body shape, gait, carried objects, vehicle association, timing, direction of travel, or other features.

This capability directly implicates one of the AIR majority’s central observations: a visually anonymous “dot” can become identifiable through context. AI can dramatically accelerate that contextual re-identification. Policy should therefore distinguish facial recognition from broader identity-resolution technologies rather than assuming that a system avoids identification simply because it does not perform face matching.

Entity resolution

Entity-resolution software links records that may refer to the same person, vehicle, address, device, or event. For example, a moving vehicle seen in aerial imagery may later appear at an ALPR camera. The plate may link to registration information. The registered address may connect to CAD calls, prior incidents, or other records. A person exiting the vehicle may then be associated with a social-media account or investigative file.

Each link can be probabilistic rather than certain. Agencies should preserve the source of each association, the confidence level, and whether a human analyst verified it. A system that collapses multiple probabilistic links into one apparently definitive identity creates both evidentiary and constitutional risk.

Behavioral analytics

AI can identify patterns such as repeated visits, unusual routes, loitering, convoy movement, recurring co-location, or deviations from an established pattern. These tools can be valuable for investigating organized crime, shootings, trafficking, vehicle theft, or coordinated violence. But behavioral analytics also move surveillance from recording what happened to inferring what the behavior means.

Those inferences may implicate association, political activity, religious practice, medical care, employment, or intimate relationships. The majority’s concern with the “whole of movements” becomes even more important when software automatically labels the meaning of those movements.

Predictive analytics

Predictive systems may use historical movement, incident, or association data to estimate where crime is likely to occur or which people or vehicles may be connected to future events. Leaders of a Beautiful Struggle does not decide the constitutionality of predictive policing. It does, however, provide a warning about the source material feeding such predictions. A prediction built from persistent movement histories may depend on data whose collection or use itself raises Fourth Amendment concerns.

Agencies should also separate intelligence prioritization from legal thresholds. A predictive score is not probable cause. A high-risk classification should not become a substitute for facts particularized to the person, place, or thing to be searched or seized.

Multimodal data fusion

Modern public-safety platforms increasingly combine video, still images, ALPR, audio, dispatch, RMS, body-worn camera metadata, geospatial records, social-media information, commercial datasets, and location information. A multimodal AI system can synthesize those inputs into a single investigative narrative.

This is precisely where system-level review becomes essential. Each underlying source may have been lawfully collected under a different legal regime. But the fused product may reveal movement, association, identity, or behavior far beyond what any one source exposed. Agencies should maintain a data-source inventory and track the legal authority, retention period, access restrictions, and permissible uses associated with each input.

Automated timeline generation

AI can now create chronological narratives automatically: vehicle appears at Location A, travels through Camera B, is detected by ALPR C, arrives near Address D, and later appears in Event E. Such timelines can materially improve investigative efficiency. They can also hide uncertainty if the system presents probabilistic connections as established facts.

Generated timelines should therefore preserve source citations, timestamps, confidence information, and analyst verification. The underlying observations should remain available for independent review by investigators, prosecutors, defense counsel, and courts.

ALPR integration

ALPR is particularly important because it can turn a visually anonymous vehicle track into an identified plate. The AIR opinion expressly discussed integration with license plate readers. Modern AI can perform that linkage automatically and at scale.

Agencies should distinguish a single ALPR lookup from analytics that reconstruct months of travel or identify all vehicles sharing a movement pattern. The latter functions move closer to the aggregation concern that animated Jones, Carpenter, Leaders of a Beautiful Struggle, and now Chatrie.

Geofence and location-data integration

A data-fusion platform may combine visual surveillance with device-location information obtained through a warrant, commercial broker, emergency process, or other source. The result can validate or extend a movement track. After Chatrie, agencies should be especially careful when precise historical device location is part of the fusion process because the Supreme Court has expressly held that government acquisition of such information can constitute a search even for a short period.

Generative AI and investigative synthesis

Generative systems can summarize hundreds of events, propose connections, identify gaps, and recommend next investigative steps. Those capabilities are useful, but they create a new risk: the system may infer a relationship that is not supported by the source record or may fail to distinguish direct observation from inference.

Investigative products should label which statements are source facts, which are algorithmic matches, and which are analytical inferences. Prosecutors should be able to trace every material factual assertion back to original evidence.

AI can create persistent surveillance from nonpersistent sensors

A city does not need one aircraft recording continuously to create a persistent movement history. AI can stitch together ALPR hits, traffic cameras, fixed cameras, body-worn camera metadata, drone footage, parking systems, access-control logs, commercial location data, and public sources. Functionally, the resulting product can resemble the retrospective record that concerned the Fourth Circuit.

That means agencies should not define “persistent surveillance” solely by the duration of one camera feed. Persistence can be created analytically through cross-system correlation.

Model opacity creates accountability problems

If software decides that two tracks belong to the same person or vehicle, the agency must understand the basis for that conclusion. Match thresholds, confidence scores, known error rates, training limitations, and analyst overrides should be documented. A vendor’s proprietary claim should not prevent the agency from understanding how an investigative inference is produced.

Human review remains essential

AI should reduce search burden, not eliminate human judgment. Analysts should independently validate significant identity, movement, or association conclusions before investigative or enforcement action. The lower the confidence or higher the consequence, the stronger the verification requirement should be.

Governance questions agencies should ask

• What raw data does the AI system ingest?
• How far back can it search?
• Can it reconstruct a person’s or vehicle’s movements across systems?
• Can it identify or re-identify anonymous tracks?
• Does it infer relationships or associations?
• What confidence or uncertainty accompanies each inference?
• Are historical searches logged and reviewable?
• Can analysts search sensitive locations or protected activity?
• Does the vendor retain prompts, queries, or agency data?
• Is agency data used to train commercial models?
• Can the system explain how it reached a material investigative conclusion?
• Are AI-generated summaries clearly distinguished from source evidence?

Implications for ALPR and Real-Time Crime Centers

ALPR

A single ALPR observation records a vehicle in public. Leaders of a Beautiful Struggle does not hold that such an observation is a search. But large historical ALPR networks can support retrospective reconstruction, pattern analysis, and reverse searches. Those functions resemble the aggregation concerns central to the AIR case.

Agencies should distinguish routine plate lookup from historical pattern-of-life analytics. Longer retention, wider geographic coverage, interstate data sharing, and integration with other systems increase the amount of movement information available from the network.

Real-Time Crime Centers

RTCCs often integrate precisely the categories discussed in the AIR opinion: cameras, ALPR, gunshot detection, dispatch, mapping, and other data. The constitutional question may therefore turn on what the center can reconstruct and how analysts use that capability.

Strong RTCC governance should define permissible query purposes, historical search thresholds, sensitive-location protections, retention, cross-system fusion, audit logs, and supervisory review.

Implications for Drones and Aerial Surveillance Programs

The case should not be read as a categorical rule against police drones. A drone deployed for a limited emergency scene, missing-person search, crash documentation, tactical overwatch, or warrant execution presents different facts from a citywide persistent surveillance database.

The closer a drone program moves toward continuous collection, broad geographic coverage, automated tracking, historical retention, and cross-system integration, however, the more relevant Leaders of a Beautiful Struggle becomes.

Key distinctions for drone policy

• Event-specific deployment versus routine persistent patrol.
• Live viewing versus searchable historical archive.
• Targeted camera field versus wide-area coverage.
• Human observation versus automated object tracking.
• Short retention versus long-term movement reconstruction.
• Standalone imagery versus integration with ALPR, facial recognition, or location databases.

Agencies developing Drone as First Responder programs should document those distinctions and periodically reassess capability as software changes. A program that is constitutionally modest at launch can become materially different after adding automated tracking, broader retention, or cross-database search.

Future Litigation: Where This Doctrine Is Likely to Matter

Leaders of a Beautiful Struggle is likely to remain important because the technologies that raise its core concern—retrospective reconstruction of movement at scale—are expanding faster than courts can decide individual cases. The following areas are particularly likely to generate litigation.

Persistent pole cameras

Federal courts remain divided over whether long-term pole-camera monitoring of a home or property becomes a search after Carpenter. Cases such as United States v. Tuggle and United States v. Moore-Bush illustrate the disagreement. Leaders of a Beautiful Struggle supplies one of the strongest appellate arguments for focusing on aggregation and retrospective capability rather than each individual image.

Automated license plate reader analytics

As ALPR retention grows and networks become geographically broader, litigants increasingly can argue that historical plate data creates a detailed movement record. Ordinary plate observation and long-term pattern-of-life analysis are not the same function. Courts may use the AIR decision when determining whether scale changes the constitutional analysis.

Real-time crime centers

RTCCs integrate multiple technologies in one analytical environment. Litigation may ask whether the relevant Fourth Amendment “search” is the original acquisition of each data point or a later fused query that reconstructs a person’s movements across many sources.

Drone as First Responder programs

DFR programs generally operate differently from AIR because flights are tied to calls for service and are often short. But expanded autonomous patrol, persistent hovering, wide-area sensors, long-term video retention, and AI tracking could move a drone program closer to the functional concerns identified by the Fourth Circuit.

Automated person and vehicle tracking

Computer vision can follow people or vehicles across a network of public cameras. Litigation may test whether a system that automatically creates a persistent track is constitutionally different from officers manually viewing individual feeds.

Commercial location data

Agencies increasingly encounter location data collected outside traditional telecom systems. After Carpenter and Chatrie, courts may scrutinize government acquisition of commercially available movement histories even when a vendor claims the data is anonymized or lawfully sold.

Connected-vehicle and telematics data

Modern vehicles generate precise location and operational data. If law enforcement gains access to historical vehicle movement through manufacturers or service providers, courts may confront whether the resulting record resembles CSLI, GPS, ALPR, or some combination of them.

Reverse-location and reverse-pattern searches

Future platforms may allow investigators to ask not only “who was here?” but “who traveled between these three locations?”, “which vehicles followed this route?”, or “which devices were repeatedly present at these events?” Such population-first queries combine the reverse-search concern of Chatrie with the aggregation concern of Leaders of a Beautiful Struggle.

AI-created pattern-of-life profiles

A system may generate a profile identifying probable home, work, associates, routines, travel corridors, and sensitive locations without any one database containing that complete picture. Courts may have to decide whether the inference itself is constitutionally significant and whether obtaining or generating it requires legal process.

Agency Policy Recommendations

Define persistent surveillance explicitly

Policy should identify systems capable of continuous, repeated, or retrospective tracking over an extended area or period. Avoid limiting the definition to one technology such as aircraft.

Require legal review before integration

Connecting two lawful systems can create a new surveillance capability. Integration projects should include legal and privacy review before deployment.

Create historical-query thresholds

Searching stored movement data should require a documented case, investigative purpose, and where appropriate a warrant or other legal process.

Limit sensitive-location analytics

Policies should restrict queries designed to identify everyone who visited a house of worship, political meeting, clinic, news organization, school, or similar sensitive location absent a strong legal basis and heightened approval.

Separate pilot data from permanent intelligence

Pilot programs should include an end-of-pilot deletion and preservation plan. Data should not silently become a permanent historical archive merely because it exists.

Publish meaningful program information

Transparency should include technology capabilities, retention, permitted crimes, data integrations, vendor role, audit process, and complaint mechanism. Describing only what a single camera can see may understate the actual system.

Instructor and Training Notes

Teaching objective 1: Sensor versus system

Ask students whether a one-pixel image of a person seems intrusive. Then add thirty-two square miles per frame, one frame per second, hours of daily coverage, forty-five days of stored imagery, ALPR, ground cameras, and retrospective analysis. The exercise illustrates why system architecture matters.

Teaching objective 2: Public movement versus aggregated movement

Use the case to distinguish the proposition that people can be observed in public from the separate proposition that government may automatically reconstruct their movements over time without constitutional limits.

Teaching objective 3: Why retention matters

Short retention reduces the window for retrospective searching. Ask students how the constitutional and policy analysis changes at 24 hours, 30 days, one year, or indefinite storage.

Teaching objective 4: Integration

Have students map the AIR system together with CitiWatch, ALPR, ShotSpotter, dispatch, and public records. Ask when separate observations become a movement history.

Teaching objective 5: Apply the reasoning to modern RTCC tools

Present a hypothetical system that combines ALPR, drone video, fixed cameras, and AI identity resolution for 90 days. Ask whether the constitutional concern is best analyzed technology by technology or as a single integrated surveillance capability.

Frequently Asked Questions

Does Leaders of a Beautiful Struggle prohibit police drones?

No. The case involved a specific citywide, persistent, retrospective aerial-surveillance program. Short-duration, event-specific drone deployments present materially different facts. The closer a drone program moves toward broad geographic coverage, automated tracking, long-term retention, and retrospective movement reconstruction, however, the more relevant the case becomes.

Does the case prohibit automated license plate readers?

No. The Fourth Circuit did not decide the constitutionality of ALPR. The opinion is relevant because AIR integrated with license plate readers and because large ALPR databases can support retrospective movement analysis. The constitutional risk generally increases as retention, geographic coverage, pattern analysis, and cross-system integration increase.

Does the case prohibit real-time crime centers?

No. RTCCs are not categorically unconstitutional. The case does suggest that agencies should evaluate what an integrated RTCC can reconstruct, how long data is retained, whether historical searches are logged, and whether multiple lawful data sources combine to reveal a detailed movement history.

Does every surveillance camera become a Fourth Amendment search?

No. The majority distinguished isolated visual observation from a persistent system capable of reconstructing the whole of a person’s movements. Ordinary cameras used for discrete observation do not automatically fall within the AIR holding.

Can law enforcement still use persistent surveillance with a warrant?

The opinion held the warrantless AIR program unconstitutional. It did not decide the precise warrant structure necessary for every comparable technology. Judicial authorization based on probable cause and particularity would materially change the analysis, but agencies would still need to define scope, duration, access, retention, and the objects of the search.

Why did low-resolution imagery still create a privacy problem?

Because identity could be inferred. A track could lead to a residence, vehicle, ground camera, license plate reader, or other identifying source. The majority focused on inferential capability rather than the number of pixels in one image.

Why was retrospective access so important?

The system let police return to stored imagery after a crime and reconstruct earlier movements. That capability removes the need to know in advance whom to follow and is one of the features that made the system resemble the historical location records at issue in Carpenter.

Does the case mean people have a complete privacy right in public movements?

No. The court did not hold that every public movement is private. It held that the aggregated and retrospective record produced by this specific system could implicate a reasonable expectation of privacy in the whole of a person’s movements.

What is the “mosaic theory”?

It is the idea that numerous individually lawful observations can collectively reveal information that is qualitatively different from any one observation. The Supreme Court has not adopted “mosaic theory” as a universal formula, but aggregation plays an important role in Jones and Carpenter, and the Fourth Circuit expressly relied on whole-of-movements reasoning here.

Does Chatrie change the analysis?

Chatrie v. United States, decided by the Supreme Court in 2026, involved a different technology—Google Location History obtained through a geofence warrant. But it reinforces the proposition that precise retrospective movement information can trigger the Fourth Amendment even for a relatively short period. Together, the cases make it harder to rely on simplistic distinctions such as “public movement” or “short duration.”

Related Cases

Carpenter v. United StatesHistorical CSLI, the limits of the third-party doctrine, and privacy in the whole of a person’s physical movements.Case Law Center page: /case-law/carpenter-v-united-states

United States v. JonesGPS monitoring, physical trespass, and the constitutional significance of prolonged tracking.Case Law Center page: /case-law/united-states-v-jones

United States v. MaynardD.C. Circuit decision developing the principle that aggregated public movements can reveal information not exposed by individual trips.Case Law Center page: /case-law/united-states-v-maynard

United States v. KnottsShort-term beeper tracking on public roads and the traditional public-movement rule.Case Law Center page: /case-law/united-states-v-knotts

United States v. KaroElectronic tracking that reveals information from inside constitutionally protected space.Case Law Center page: /case-law/united-states-v-karo

Kyllo v. United StatesTechnology-assisted acquisition of information about the home and limits on sense-enhancing technology.Case Law Center page: /case-law/kyllo-v-united-states

Riley v. CaliforniaDigital-device privacy and the Supreme Court’s recognition that modern digital technologies differ qualitatively from traditional physical objects.Case Law Center page: /case-law/riley-v-california

Chatrie v. United StatesSupreme Court decision holding that acquisition of precise historical Google Location History is a search even for a limited period.Case Law Center page: /case-law/chatrie-v-united-states

United States v. TuggleLong-term pole-camera surveillance and competing approaches to aggregation after Carpenter.Case Law Center page: /case-law/united-states-v-tuggle

United States v. Moore-BushResidential pole cameras, prolonged surveillance, and a fractured en banc treatment of Carpenter and Jones.Case Law Center page: /case-law/united-states-v-moore-bush

United States v. HoustonExtended pole-camera observation and the treatment of activities exposed to public view.Case Law Center page: /case-law/united-states-v-houston

United States v. HayVideo surveillance and the relationship between public exposure, duration, and Fourth Amendment protection.Case Law Center page: /case-law/united-states-v-hay

California v. CiraoloWarrantless fixed-wing aerial observation of a residential curtilage from navigable airspace.Case Law Center page: /case-law/california-v-ciraolo

Florida v. RileyHelicopter observation of residential curtilage from lawful navigable airspace.Case Law Center page: /case-law/florida-v-riley

Dow Chemical Co. v. United StatesAerial photography of an industrial complex and the distinction between open industrial areas and the home.Case Law Center page: /case-law/dow-chemical-v-united-states

California v. GreenwoodExposure to the public and the limits of reasonable expectations of privacy; useful contrast to modern aggregation cases.Case Law Center page: /case-law/california-v-greenwood

Primary Authorities and Research Sources

Leaders of a Beautiful Struggle v. Baltimore Police Department, 2 F.4th 330 (4th Cir. 2021) (en banc)Official Fourth Circuit en banc opinion, including majority, concurrences, and dissents.

Leaders of a Beautiful Struggle v. Baltimore Police Department, 979 F.3d 219 (4th Cir. 2020)Original panel decision, later reheard en banc.

Accessible En Banc Opinion TextJustia version of the Fourth Circuit decision.

Carpenter v. United StatesHistorical CSLI and digital location privacy.

United States v. JonesGPS tracking and prolonged surveillance.

United States v. MaynardD.C. Circuit aggregation analysis later discussed in Jones.