By: Simon Caine*
Abstract.
This Article examines how AI-assisted drone programs—exemplified by Montgomery County, Maryland’s Drone as First Responder initiative—are rapidly outpacing the constitutional doctrines meant to restrain government surveillance. As autonomous functions, intelligent tracking, and cloud-based video archives migrate from military systems into domestic policing, traditional Fourth Amendment frameworks built around human perception and episodic observation no longer suffice. Supreme Court precedents on aerial views, sense-enhancing technologies, and long-term tracking—Ciraolo, Riley, Kyllo, Jones, and Carpenter—offer partial guidance, but none anticipated algorithmic reacquisition, supervised autonomy, or the mosaic-building power of persistent archives. The Fourth Circuit’s en banc decision in Leaders of a Beautiful Struggle v. Baltimore Police Department underscores the constitutional stakes: access to a multi-week aerial archive is itself a search requiring judicial oversight.
Against this doctrinal backdrop, this Article argues that Maryland stands at a pivotal moment. The DFR program’s rapid-response value is undeniable, yet its technical architecture—semi-autonomous routing, AI-assisted tracking, high-resolution imaging, and cloud retention—creates the latent capacity for persistent, programmatic surveillance. The constitutional question, therefore, shifts from isolated drone flights to system design: who or what initiates surveillance; how long data is retained; when software may track, reacquire, or analyze; and how proximity to the home triggers enhanced protections.
To address these challenges, this Article proposes a comprehensive Maryland statute modeled on, but substantially updating, Illinois’s Freedom from Drone Surveillance Act. The statute would impose a warrant-first regime for investigative uses; bar facial recognition and appearance-based re-identification absent judicial authorization; codify curtilage-specific minimization; establish audit-logging, retention, and secondary-use limits; and require human-in-the-loop confirmation for AI-triggered tracking and analytic queries. It further integrates reliability, cybersecurity, and FAA-mandated safety considerations into the constitutional concept of “reasonableness,” recognizing that flawed autonomy or unvalidated analytics can render state surveillance unlawful in both design and effect.
Ultimately, this Article contends that Maryland must legislate before autonomy becomes entrenched and archives become default. By embedding constitutional values into technical parameters—rather than relying solely on post-hoc adjudication—the state can ensure that AI-enabled policing enhances public safety without eroding fundamental privacy protections. When drones begin to decide, they must do so within a constitutional framework deliberately built for the machine age.
- Introduction.
Two days, two podiums, one warning: artificial intelligence’s (“AI”) speed is outpacing both reliability and the law.[1] In September 2025, two headlines told the same uneasy story.[2] First, senior Air Force officials publicly acknowledged the limits of contemporary AI including accuracy problems, bias, “hallucinations,”[3] and the extensive human oversight and data infrastructure required to keep these tools safe and useful.[4] At the Air & Space Forces Association’s 2025 Air, Space & Cyber Conference, it was man versus machine.[5] The task was to draft a course of action (“COA”) on how to proceed with a complex issue.[6] The AI system generated a COA in roughly ten seconds, compared to about sixteen minutes for a human operator; yet, as Maj. Gen. Robert Claude, the Space Force representative to the Advanced Battle Management System (“ABMS”) Cross-Functional Team, cautioned that the resulting COAs “weren’t necessarily completely viable.”[7]
The next day at the United Nations, Ukrainian President Volodymyr Zelenskyy warned that AI-accelerated warfare, including drones, was propelling the world toward the most destructive arms race in human history.[8] He predicted that in the near term, unmanned aerial systems will likely engage one another in combat, conduct operations against critical infrastructure, and select and strike human targets without a human operator in the decision loop—save for limited oversight by those managing the underlying AI systems—and, on that basis, he called for the establishment of internationally binding rules governing the development and use of AI-enabled weapons.[9] Those global warnings of the future of AI matter locally: the same classes of autonomy, perception, and decision support migrate rapidly from battlefields to domestic public‑safety systems.[10]
In Maryland, this fusion of AI-powered drones is no abstraction.[11] Montgomery County’s “Drone as First Responder” (“DFR”) program now provides rapid aerial response to 911 calls, and the county publicly reports its activity.[12] In September 2025 alone, the department logged 198 flights on its dashboard.[13] In its first year, the program responded to well over 1600 calls and was frequently first on scene—often within one minute.[14] “Quicker situational awareness and faster response times are meaningful gains that can, in turn, reduce risk, permit call cancellations, and generate probative video in contested cases.”.[15]
However, scale and machine assistance change the constitutional question.[16] The Supreme Court’s aerial surveillance decisions, namely California v. Ciraolo and Florida v. Riley, permit brief, naked-eye observations from lawful airspace, with Dow Chemical Co. v. United States addressing the separate issue of imaging industrial complexes.[17] The Court later clarified this doctrine in Kyllo v. United States, establishing that using sense-enhancing technology to monitor a home constitutes a search.[18] More recently, the line of cases from United States v. Knotts to Carpenter v. United States shows the Court’s growing concern with long-term, technology-enabled tracking.[19] Most salient for Maryland, the Fourth Circuit’s en banc decision in Leaders of a Beautiful Struggle v. Baltimore Police Dep’t held that access to a multi‑week aerial archive is itself a search requiring a warrant.[20] AI‑assisted drone operations sit at the intersection of these lines: what begins as incident‑tethered overflight can, without guardrails, drift toward persistent, automated surveillance.[21]
This Article contends that Maryland should not leave AI‑enabled drone operations to general Fourth Amendment doctrine and ad hoc departmental policy.[22] The General Assembly should enact a comprehensive statute, modeled on Illinois’s framework; yet expressly tailored to govern drones and their use of AI, and likely autonomy.[23] Such an act would adopt a warrant‑first rule for investigative uses with narrow, time‑limited exceptions (exigency, missing persons, disasters, scene documentation); prohibit facial recognition and person re‑identification on drone‑gathered imagery absent a warrant and minimization order; cabin thermal use around homes and curtilage; require human‑in‑the‑loop confirmations for automated tracking; specify retention, secondary‑use, and audit‑log duties for video and coordinate metadata; and provide exclusionary and civil remedies.[24]
The harder question—and the title’s provocation—is what happens when drones begin to decide for themselves.[25] As supervised autonomy replaces moment-to-moment human input, constitutional reasonableness must account for programmatic choices regarding delegating sensing, target selection, and routing to software, as well as the retrospective power of cloud archives.[26] Part II situates Maryland’s Drone as First Responder program against this backdrop, framing the core constitutional problem and the Article’s contributions.[27] Part III then translates contemporary AI-assisted capabilities into concrete surveillance powers and reliability risks.[28] Part IV maps those powers onto existing doctrine, from the aerial-observation and sense-enhancement cases to the long-term tracking and mosaic line.[29] Part V explains why autonomy and archive queries press toward warrants, minimization, and heightened protections near the home.[30] Part VI advances a Maryland statute, adapted from Illinois’s Freedom from Drone Surveillance Act and expressly tailored to AI and autonomy.[31] Part VII integrates safety, cyber- and navigation-reliability, and institutional practice into the reasonableness analysis,[32] and Part VIII concludes by situating Maryland’s model within the broader national project of designing constitutional limits for the machine age.[33]
Moreover, this Article makes two contributions.[34] First, it reframes contemporary drone deployments—particularly those employing Smart Track, PinPoint, and mission replays—as legally distinct from traditional overflight or fixed-camera surveillance.[35] These tools are not passive recorders.[36] They perceive, decide, and act with measurable independence, effectively transforming aerial observation into an AI-mediated investigative process.[37] As integration with cloud orchestration systems such as DJI’s FlightHub 2 matures, the shift from operator aid to supervised autonomy will accelerate, further blurring the boundary between human oversight and machine initiative.[38]
Second, this Article advances a legislative blueprint.[39] Drawing from Illinois’s amended Freedom from Drone Surveillance Act, it proposes a Maryland statute that: (1) requires warrants for any drone operation involving thermal imaging near a home, person tracking, or archive querying; (2) bars facial recognition and person re‑identification absent judicial approval and minimization; (3) mandates real‑time flagging of AI activations and retention of override logs; and (4) provides both exclusionary and civil remedies for violations.[40] Together, these measures would align state practice with constitutional principle, ensuring that AI-enhanced policing remains bounded by law rather than by technical capacity.[41]
II. Framing the Problem and the Contribution.
This Part narrows the lens from global warnings about AI-accelerated warfare to Maryland’s concrete experiment with AI-assisted policing.[42] Specifically, it traces how Montgomery County’s DFR program already embodies the fusion of autonomy, rapid aerial response, and cloud-based archives, and it shows why those operational choices strain existing Fourth Amendment doctrine and Article 26.[43] Furthermore, this Part outlines the dual challenge facing Maryland, doctrinal and institutional, and distills the Article’s core contributions: reconceptualizing AI-enabled drone deployments as qualitatively different from traditional aerial surveillance, and sketching a legislative blueprint to govern autonomy, archives, and analytics before they harden into unchecked practice.[44]
Today, Maryland stands at the legal edge of a technological transformation.[45] Police drones are no longer speculative or experimental—they are operational, AI-assisted, and increasingly semi-autonomous.[46] Since November 2023, the Montgomery County Police Department (“MCPD”) has deployed its DFR program using the DJI Matrice 350 RTK equipped with Zenmuse H20-series sensors.[47] In the month of September 2025 alone, the MCPD logged 198 DFR flights, publicly reported on a live dashboard that catalogs deployments, mission types, and response times.[48] The DFR program’s utility is unquestionable: it offers near-instant overhead situational awareness, automates target tracking, and delivers probative video in contested environments.[49] As a matter of operational design, the DFR program reduces response time, allows for rapid call triage, and when coupled with analytics provides an evidentiary and geographic context that traditional patrols cannot match.[50]
Yet technological capacity does not equate to constitutional compliance.[51] The same features that make these systems efficient also make them constitutionally fraught.[52] Automated person tracking (“Smart Track”), waypoint mission replay, and persistent archival mapping (“PinPoint”) collectively generate surveillance capabilities that exceed the contemplation of earlier Fourth Amendment jurisprudence.[53] The doctrinal lineage running from California v. Ciraolo and Florida v. Riley to Dow Chemical Co. v. United States tolerated brief, naked-eye overflights from lawful airspace.[54] Those precedents presumed episodic human observation such as fleeting glances from the sky, not continuous, automated monitoring.[55] They did not imagine algorithmic reacquisition, multi-day replay, or machine learning models trained to detect, track, and catalog patterns of human movement.[56]
The Supreme Court has begun to address the constitutional limits of advanced surveillance, signaling a potential framework for analyzing algorithmic reacquisition.[57] Modern cases such as United States v. Jones, where location-tracking technology crossed the line from merely augmenting to impermissibly enhancing and allowing police to use a GPS-tracking device to remotely monitor and record a vehicle’s movements over 28 days, and Carpenter v. United States, held that a person does not surrender all Fourth Amendment protection by venturing into the public sphere, do imagine a somewhat more advanced algorithmic reacquisition.[58] These later decisions recognize that aggregation itself, the persistent tracking and data-driven reconstruction of life’s patterns, can constitute a search.[59]
Most recently, the Fourth Circuit’s en banc decision in Leaders of a Beautiful Struggle v. Baltimore Police Department applied that logic to Maryland, holding that access to a multi‑week aerial archive opened “an intimate window” into a person’s associations and activities, and thereby violated the reasonable expectation of privacy individuals have in the whole of their movements.[60] The opinion serves as both a constitutional compass and a cautionary tale: it signals that Maryland’s tolerance for AI-enhanced aerial surveillance cannot rest on federal precedent that never contemplated autonomy.[61]
The policy posture mirrors the doctrinal one.[62] MCPD’s current rules limit DFR launches to 911 calls or officer requests; prohibit audio recording and facial recognition; and require that drone video be managed like body‑worn camera footage.[63] These commitments demonstrate commendable restraint but still remain internal, voluntary, and reversible, as they are not law.[64] No Maryland statute presently codifies limits on autonomous sensing, mandates human‑in‑the‑loop confirmation before initiating AI tracking, or regulates the retention and secondary use of coordinate metadata.[65] This statutory vacuum is not trivial.[66] As autonomy expands, the locus of governmental intent shifts from discrete officer decisions to system configurations.[67] When software, in place of an officer, decides when to begin, continue, or modify surveillance, the Fourth Amendment’s traditional focus on an individual’s intent to search must evolve.[68] In this environment, constitutional reasonableness depends as much on programmatic design as on officer conduct.[69] Maryland’s challenge, therefore, is dual: doctrinal and institutional.[70]
Doctrinally, courts must adapt the “search” inquiry to address AI-driven perception and decision‑making, acknowledging that persistent algorithmic tracking constitutes a search in both effect and design.[71] Institutionally, lawmakers must move from reactive adjudication to proactive governance, ensuring that autonomy operates under law rather than precedent-by-analogy.[72] As the Fourth Circuit’s decision in Leaders illustrates, waiting for litigation to define constitutional limits invites after‑the‑fact remedies for harms that are technologically irreversible.[73]
If Maryland fails to act while the DFR program remains in its formative stage—before autonomy becomes default and data archives self‑referential—it risks the same reactive posture that has characterized much of Fourth Amendment adaptation in the digital era.[74] By legislating early, Maryland can establish durable norms of transparency, warrant accountability, and privacy that outlast particular vendors and platforms.[75] A statute designed for autonomy would not impede responsible drone use; it would enable it, grounding operational legitimacy in law and public trust rather than discretionary policy.[76]
Having framed the constitutional and statutory stakes of Maryland’s emerging drone infrastructure, the next task is to translate those abstractions into technical and operational realities.[77] The constitutional argument cannot be understood apart from the hardware, software, and data pipelines that define how these systems actually function.[78] What the law calls a “search” depends on what the machine is engineered to perceive, record, and decide.[79] Part III, therefore, grounds the discussion in the concrete mechanics of MCPD’s DFR program—the platform’s architecture, its built‑in intelligent functions, and the reliability and safety risks that accompany its increasing autonomy.[80] By tracing the progression from human‑directed assistance to supervised autonomy, Part III establishes the technological foundation for the doctrinal and legislative analysis that follows.[81]
III. Contemporary AI‑Assisted Capabilities: Concrete Powers and Reliability Risks.
Having established Maryland’s doctrinal and statutory gap, this Part turns from principle to mechanism; from what the law must address to what it must regulate.[82] Constitutional reasonableness and statutory design hinge on the technical details of how the MCPD’s DFR program actually works.[83] The platform’s architecture, hardware, onboard intelligence, and network integration, defines both the reach and risk of police aerial operations.[84] Understanding autonomy, analytics, and sensing functions thus becomes a prerequisite to evaluating whether a drone’s behavior constitutes a “search,” who performs it, and when it becomes persistent enough to invoke the Fourth Amendment.[85]
- Autonomy (In‑the‑Loop / On‑the‑Loop / Out‑of‑the‑Loop).
“Autonomy” in public‑safety aviation denotes the capacity of a system to perceive its environment, make decisions, and execute actions with reduced or no continuous human control.[86] Operationally, agencies distinguish three modes: (1) human‑in‑the‑loop, in which a pilot must approve each action; (2) human‑on‑the‑loop, where the system acts subject to human supervision and override; and (3) human‑out‑of‑the‑loop, in which humans review outcomes only after the fact.[87]
Maryland’s DFR deployments exemplify the middle ground.[88] Using DJI Pilot 2 and FlightHub 2, operators can pre‑program missions, monitor them in real time, and intervene if needed, but most routine functions, from waypoint replay to return‑to‑home, proceed automatically.[89] Built‑in obstacle avoidance, intelligent landing, and route optimization further minimize pilot input.[90] Although the Federal Aviation Administration (“FAA”) still requires a certificated Remote Pilot in Command to retain ultimate authority, the technical reality is that algorithmic assists increasingly determine what the drone perceives and records.[91]
Autonomy also encodes intent.[92] Policy choices about autonomy modes, geofences, altitude caps, dwell limits, and object‑tracking thresholds effectively program investigative scope and duration.[93] When the system itself executes a search behavior, such as reacquiring a tracked vehicle, orbiting a coordinate, or logging metadata without a fresh human command, the constitutional act is embedded in design.[94] The traditional focus on an officer’s subjective “intent to search” must therefore evolve to include the agency’s configuration of its autonomous systems.[95]
B. Predictive Analytics (Detection, Re‑Identification, Trajectory, Anomaly).
Beyond flight control, modern police drones perform analytics once limited to forensic labs.[96] “Predictive analytics” describes a pipeline that ingests live video and sensor data, extracts features, runs trained models, and outputs inferences, including object detection, appearance re‑identification, motion forecasting, or anomaly alerts.[97] On the Matrice 350 RTK platform (the current drone model used for the DRP), these capabilities operate both at the edge, through the camera payload and in the cloud and through DJI’s Mobile SDK and Cloud API.[98] Each analytic layer introduces governance variables: model versioning, decision thresholds, false‑positive and false‑negative rates, and human‑override logging.[99] Bias in computer‑vision systems is well documented; the National Institute of Standards and Technology’s (“NIST”) Face Recognition Vendor Test series shows demographic differentials even in top‑performing algorithms.[100] Unvalidated analytics risk disproportionate surveillance of particular populations or environments.[101]
Illinois’s Freedom from Drone Surveillance Act reflects this concern by restricting identity‑based analytics and requiring reporting and retention controls adaptable to any model‑driven workflow.[102] For Maryland, the constitutional implication is clear: predictive analytics shatters the temporal gap between observation and inference.[103] When software flags an individual or a vehicle based on probabilistic models, the “decision to search” may already have occurred before any officer consciously intervenes.[104] These concerns, however, are not merely theoretical with recent Maryland arrests demonstrating how algorithmic surveillance systems already operate beyond traditional constitutional constraints.[105] Specifically, two recent Maryland cases demonstrate exactly how predictive analytics collapse the temporal gap between observation and inference in real-world scenarios.[106] The first is the Silver Spring Stabbing Case.[107] In this Montgomery County case, the drone footage enabled a conviction “almost exclusively based on drone footage” without calling the victim as a witness.[108] The drone captured the stabbing and then followed the perpetrator as he walked away, with police apprehending him minutes later.[109] This exemplifies how the decision to search occurred algorithmically—the drone’s tracking capabilities automatically flagged and followed the suspect before any conscious officer intervention in the arrest decision.[110]
The second case is the Harford County Juveniles Case.[111] Here, the police used thermal imaging technology that demonstrates even more sophisticated algorithmic inference.[112] The drone’s heat vision pinpointed the suspects hiding in dense woods, with the technology making the identification decision before officers could visually locate them.[113] Deputies found the juveniles had broken into at least twenty vehicles and possessed a loaded firearm.[114] Here, the algorithmic analysis of heat signatures determined the search targets before human officers made conscious decisions about where to look.[115] Both cases illustrate how predictive policing does not merely detect crime but actively codes it into existence, with algorithmic suspicion becoming the basis for preemptive action.[116] DFR operates from a command center in Rockville, with drones tracking suspects and returning to rooftop stations to await the next 911 call, showing the automated, always-ready nature of these systems.[117] These examples underscore the need for Maryland’s statutory governance to extend to algorithmic inference itself, thereby requiring transparency, validation, and auditability of the analytic models used in these law-enforcement flight systems.[118]
C. Facial Recognition (1:1 Verification vs. 1:N Identification).
Facial recognition extends the analytic continuum from detection to identity.[119] Technically, it entails (1) detecting a face; (2) extracting a numerical embedding; and (3) comparing that vector to one or more stored templates.[120] One-to-one verification tests a claimed identity, defined as matching a suspect to a mugshot, whereas one-to-many identification searches a gallery for any match.[121] Performance metrics, such as false-accept and false-reject-rates, vary with demographic composition and imaging conditions.[122] Mounted on drones, facial recognition multiplies both reach and risk.[123] The combination of aerial scale, long-range optics, and continuous recording enables identity-at-scale: a capability to re-identify individuals across multiple scenes and times.[124] Illinois, for example, prohibits equipping public-safety drones with facial-recognition or weapon systems absent express statutory authorization and narrow exceptions.[125] That prohibition reflects a constitutional instinct toward categorical rules that once identification becomes dragnet-capable the particularized suspicion evaporates.[126]
Maryland has taken a narrower but significant step on facial-recognition transparency outside the drone context.[127] In 2024, the General Assembly enacted Chapter 808, creating a new Criminal Procedure subtitle on Facial Recognition Technology that, among other safeguards, requires the State to disclose in criminal and juvenile discovery: whether facial recognition technology was used in an investigation, the name of each system employed, the databases searched, and the results that led to further investigative action.[128] That statutory disclosure duty recognizes identity analytics as a distinct, discovery-worthy investigative technique.[129] If Maryland contemplates any biometric analytics in law-enforcement drone programs, it should build on that foundation and adopt a warrant-first regime that treats facial or appearance re-identification as a separate search event requiring judicial authorization, minimization, and human verification.[130] Additionally, gallery composition, model accuracy, and deletion protocols must be judicially reviewable.[131] Without such limits, AI-driven identification risks transforming every flight into a general search for identity rather than evidence.[132]
D. Tech → Legal Issue Map (Capabilities to Legal Triggers).
Each capability described above triggers a distinct constitutional concern.[133] First, autonomy shifts the inquiry from momentary officer conduct to programmatic design, making autonomy-mode selection and safety envelopes relevant to whether a search is “reasonable” and “particularized.”[134] Second, persistence, which is enabled by repeated sorties, data retention, and analytics, renders aggregated observation functionally equivalent to long-term tracking, invoking Carpenter v. United States and Leaders of a Beautiful Struggle v. Baltimore Police Department.[135] Third, sense enhancement through thermal or high-gain optical zoom triggers Kyllo v. United States when directed toward homes and curtilage, requiring warrants and minimization.[136] Finally, because the FAA governs flight safety but not privacy, states remain free to legislate purpose, retention, transparency, and biometric limits, including discovery regimes like Maryland’s facial-recognition statute, without intruding on navigable-airspace federal regulation.[137] The result is a complex, but navigable, map linking technical affordances to legal triggers.[138] By codifying these relationships through warrants, narrow exceptions, analytic-use restrictions, deletion schedules, and reporting requirements, Maryland can operationalize constitutional values in statutory form, rather than relying on ex post judicial correction.[139]
IV. Constitutional Frame: From Glances to Mosaics.
Part III established the technological foundations of autonomy, analytics, and sensing.[140] This Part turns to the doctrine that must govern them.[141] The Supreme Court’s precedents on aerial observation, sense enhancement, and long-term tracking were built for human perception in discrete moments, not for continuous, algorithmic surveillance.[142] Yet those cases provide the constitutional scaffolding from which Maryland must now legislate and adjudicate.[143] The doctrinal story runs from brief, naked-eye glances in Ciraolo, Riley, and Dow Chemical; through the home-protective logic of Kyllo; and culminates in the mosaic theory of Jones, Carpenter, and Leaders of a Beautiful Struggle.[144] Together, these cases chart a movement from observation to aggregation and momentary oversight to programmatic surveillance—the constitutional frame within which AI‑assisted drones must now be judged.[145]
- Aerial Observation Precedents: Ciraolo, Riley, Dow Chemical.
The Court’s early aerial‑surveillance cases tolerated warrantless observation from lawful airspace.[146] In California v. Ciraolo, the police flew a private aircraft over a fenced backyard at 1,000 feet, spotted marijuana plants with the naked eye, and used those observations to secure a warrant.[147] The Court held that because the officers were in navigable airspace and used only unaided vision, the defendant’s expectation of privacy was not one that society would recognize as reasonable.[148] Similarly, in Florida v. Riley, the Court found no search when an officer observed the interior of a greenhouse from a helicopter 400 feet above ground, emphasizing that the vantage point was lawful and that any member of the public could have seen the same thing.[149] Finally, in Dow Chemical Co. v. United States, the Court permitted the Environmental Protection Agency’s use of aerial mapping cameras to photograph an industrial complex, distinguishing between the home’s intimate curtilage and a commercial facility’s open areas.[150] Together, these cases created a rule of visual neutrality: what is seen from lawful airspace with the naked eye is not a search.[151]
However, that rule assumed a fleeting, human glance.[152] None of these cases envisioned continuous imaging, data retention, or algorithmic enhancement.[153] They also predated the commercial availability of high‑zoom optics and networked drones.[154] The doctrinal logic that treats brief observation from public vantage as constitutionally benign becomes unstable when those observations persist, replay, and recombine through AI‑assisted systems.[155] What begins as Ciraolo’s one‑off flyover risks maturing into a standing aerial presence that continuously documents human life below.[156]
B. Sense Enhancement and the Home: Kyllo and the Warrant Threshold.
The constitutional boundary between permissible observation and search becomes most salient at the home.[157] In Kyllo v. United States, the Court held that when the government uses a device not “in general public use” to explore details of a home otherwise unknowable without physical intrusion, it conducts a search.[158] There, officers used a thermal imager to detect heat patterns emanating from a private residence suspected of housing a marijuana grow operation.[159] Specifically, in 1991, Agent William Elliott of the United States Department of the Interior suspected that marijuana was being grown in Danny Kyllo’s home, part of a triplex in Florence, Oregon.[160] Since indoor marijuana growth typically requires high-intensity lamps, at 3:20 a.m. on January 16, 1992, Agents Elliott and Dan Haas used an Agema Thermovision 210 thermal imager to scan the triplex.[161] Thermal imagers detect infrared radiation, which virtually all objects emit but which is not visible to the naked eye.[162] The scan showed that Kyllo’s garage roof and a side wall were relatively hot compared to the rest of his home and substantially warmer than the neighboring units.[163] Based in part on the thermal imaging, a Federal Magistrate Judge issued a warrant to search Kyllo’s home, where the agents found marijuana growing.[164]
The Court reasoned that the use of thermal imaging technology to detect heat emanating from a home constitutes a search under the Fourth Amendment because it intrudes upon the privacy of the home, a constitutionally protected area. [165] The Court emphasized that the Fourth Amendment’s protection of the home is not contingent on the type or amount of information obtained.[166] Instead, it is the act of obtaining information about the interior of the home through sense-enhancing technology, which could not otherwise be obtained without physical intrusion and constitutes a search.[167] The Court further held that employing thermal imaging technology, then not commonly available to the public, intruded upon a reasonable expectation of privacy.[168] The ruling emphasized safeguarding the home from governmental surveillance as technology advances, ensuring the Fourth Amendment’s protections remain strong against emerging monitoring techniques.[169] This principle applies particularly when the technology used is not in general public use, as it ensures the preservation of privacy against government intrusion that existed at the time the Fourth Amendment was adopted.[170]
Additionally, the Court rejected the government’s argument that thermal imaging was constitutional because it did not detect private activities or intimate details.[171] The Court clarified that all details within the home are considered intimate, as the home is held safe from prying government eyes.[172] It further noted that allowing such surveillance without a warrant would leave homeowners vulnerable to advancing technology, eroding the privacy guaranteed by the Fourth Amendment.[173] The Kyllo decision underscored that the use of such technology to explore details of the home, even if it only reveals heat patterns, crosses the threshold of privacy and is presumptively unreasonable without a warrant.[174]
Lastly, the Court highlighted that the Fourth Amendment’s protection extends beyond physical intrusion, encompassing technological surveillance that reveals details of the home.[175] This approach aligns with the principle that the Fourth Amendment protects against both physical and technological encroachments on privacy.[176] The Court’s reasoning reflects a broader concern about the impact of advancing technology on privacy rights, emphasizing the need to adapt constitutional protections to modern contexts.[177]
Police drones equipped with high‑resolution zoom, infrared, or thermal sensors squarely implicate Kyllo.[178] Even if commercially available, such tools are not “in general public use” when mounted on government platforms for investigative purposes.[179] When a drone lingers over a residence, scanning through foliage or detecting heat signatures, it converts observation into search.[180] Under Kyllo, the constitutional test is not the altitude of the aircraft but the intimacy of the information obtained.[181] In Maryland, Article 26 of the Declaration of Rights is interpreted consistently with, and sometimes more broadly than, the Fourth Amendment and amplifies that protection.[182]
C. Long‑Term Tracking and Aggregation: Knotts, Karo, Jones, Carpenter, Leaders of a Beautiful Struggle.
The next doctrinal evolution concerns duration and aggregation.[183] The first set of tracking cases involved beepers to monitor a suspect; in those cases, the courts distinguished between observation of public travel and intrusion into private space.[184] In United States v. Knotts, the Court upheld the limited use of a beeper transmitter to track a vehicle along public roads, reasoning that the movements were visible to anyone.[185] The Court emphasized that a person traveling in an automobile on public thoroughfares has no reasonable expectation of privacy in his movements from one place to another, as those movements are “voluntarily conveyed to anyone who wanted to look.”[186] The beeper merely augmented police efficiency, enabling officers to follow the vehicle more easily, but did not expose any private facts or information otherwise shielded from public view.[187] Crucially, the Court declined to address whether “dragnet-type law enforcement practices” involving prolonged or comprehensive electronic tracking might implicate different constitutional concerns, expressly reserving that question for another day.[188]
Conversely, one Term later, in United States v. Karo, the Court held that monitoring a beeper inside a private residence violated the Fourth Amendment because it revealed information not otherwise observable.[189] In Karo, the Court drew a constitutional line when beeper monitoring disclosed information from within a private residence, holding that such monitoring constituted a search requiring a warrant.[190] Although the initial installation of the beeper in a container with the original owner’s consent did not violate the Fourth Amendment, the subsequent monitoring revealed facts, specifically the presence of the container inside a home, that could not have been obtained through visual surveillance from a lawful vantage point.[191] The Court underscored the heightened constitutional protection afforded to the home, reasoning that electronic monitoring which substitutes for physical intrusion impermissibly invades a reasonable expectation of privacy.[192] Unlike Knotts, where monitoring tracked public movements only, Karo established that sense-enhancing technology becomes constitutionally significant when it exposes information about the interior of the home that would otherwise remain unknowable without entry.[193]
Three decades later, United States v. Jones confronted GPS tracking.[194] The majority found that attaching and monitoring a GPS device on a suspect’s car for 28 days constituted a search.[195] Writing for the majority, Justice Scalia revived a property-based understanding of the Fourth Amendment and reasoned that the physical trespass on Jones’s vehicle, an “effect” enumerated in the Amendment, combined with the purpose of obtaining information, was sufficient enough to trigger constitutional scrutiny.[196] The Court distinguished earlier beeper cases such as Knotts and Karo on the ground that they involved no physical occupation of private property by the government.[197] While the majority resolved the case on trespass grounds, five Justices separately emphasized that long-term, continuous tracking impinges on reasonable expectations of privacy by enabling the government to assemble a comprehensive record of a person’s public movements, raising concerns about aggregation, duration, and the qualitative difference between short-term observation and prolonged surveillance.[198] This Fourth Amendment rationale further matured in Carpenter v. United States.[199] In Carpenter, the Court held that accessing 127 days of historical cell‑site location data was a search because it created “a detailed chronicle of a person’s physical presence.”[200] The Court recognized that technological persistence transforms what might once have been fleeting public exposure into comprehensive surveillance.[201] Applied to drones, the Carpenter holding is direct: a mosaic of repeated or automated overflights can reveal more about a person than any single flight viewed in isolation.[202]
- Mapping AI‑Drone Functions to Doctrine: From Discrete Overflights to Continuous Algorithmic Observation.
The doctrinal arc from Ciraolo to Leaders converges on a simple truth: persistence and enhancement redefine the constitutional event.[203] AI‑enabled drones collapse the distinctions among the precedents.[204] They combine Ciraolo’s lawful vantage with Kyllo’s sense‑enhancement capacity and Carpenter’s aggregation power.[205] When autonomy allows a drone to reacquire a target automatically, or when predictive analytics mine stored footage to trace movements, the resulting surveillance is no longer “plain view”—it is algorithmic observation with constitutional consequences.[206]
The Fourth Circuit’s en banc decision in Leaders of a Beautiful Struggle v. Baltimore Police Department illustrates how persistent aerial surveillance can implicate Fourth Amendment protections.[207] In that case, the Baltimore Police Department’s Aerial Investigation Research (“AIR”) program, which flew planes over 90% of the city during daylight hours, was deemed an unconstitutional “search” because it enabled retrospective tracking of individuals across space and time, revealing behavioral patterns and associations otherwise unknowable without technology.[208] Drawing on the Supreme Court’s mosaic theory articulated in Carpenter v. United States and United States v. Jones, the court emphasized that aggregated surveillance of public movement can violate reasonable expectations of privacy when it enables law enforcement to construct “an intimate window into a person’s life.”[209]
MCPD’s DFR program, however, differs significantly from the AIR program; in Leaders: drones are launched reactively to specific 911 calls, remain airborne for only minutes per incident, and do not continuously blanket geographic areas or record round-the-clock coverage.[210] Still, DFR missions are logged and video is retained for evidentiary review, and depending on how such data is combined or reused, even individually innocuous deployments could collectively generate a privacy-invasive surveillance mosaic over time.[211] Unlike AIR, DFR surveillance is narrower and event-driven, more analogous to the short-term aerial flyovers upheld in California v. Ciraolo and Florida v. Riley.[212]
However, as the Leaders majority held, the constitutional line is not merely technological capability but what law enforcement deduces from the aggregate data.[213] Therefore, DFR’s use of zoom, thermal imaging, and object tracking tools, especially when used near residential areas, raises questions about curtilage and incidental capture, even if policies exist to minimize such exposures.[214]
In this doctrinal gray zone between permissibly brief public observation and constitutionally suspect aggregate tracking, DFR’s legality cannot rest on analogy alone.[215] There is a compelling need for legislative action to establish clearer statutory guardrails: defining acceptable flight parameters, limiting data retention, clarifying public reporting, and institutionalizing oversight mechanisms to ensure that police drone programs do not drift into constitutionally impermissible mass surveillance.[216]
V. Autonomy, Archives, and the Case for Warrants.
The doctrinal trajectory traced above demonstrates both the strength and the limits of constitutional adjudication.[217] Ciraolo, Kyllo, Carpenter, and Leaders of a Beautiful Struggle together delineate when aerial surveillance becomes a “search,” but they do not specify how an autonomous or AI‑assisted system must be designed, monitored, or audited to remain lawful.[218] The Fourth Amendment sets boundaries ex post—after injury, litigation, or precedent slowly adapts.[219] Legislatures, by contrast, can operationalize those principles ex ante,[220] by embedding warrant procedures, retention rules, and oversight mechanisms directly into statutory code.[221] Furthermore, the legislative process can be proactive, instead of the reactive process that is required by leaving these issues to the judiciary.[222] Having established the constitutional limits on surveillance, this Part turns from doctrine to design.[223] Specifically, the following Part advocates for the Legislature to take a proactive approach to protect the privacy of its citizens before they are further eroded.[224]
The challenge for Maryland lies not merely in recognizing when AI-enabled observation becomes a “search,” but in constructing procedural safeguards that prevent unlawful searches from occurring in the first place.[225] The task is to convert the Fourth Amendment’s retrospective remedies into ex ante controls—rules governing how autonomy is delegated, how archives are accessed, and how proximity to the home triggers heightened protections.[226] Together, these parameters define a warrant-based architecture for lawful drone use in the age of artificial intelligence.[227]
- When Drones Decide: Delegation from Human to Software.
Autonomy fundamentally reconfigures the question of “who” conducts a search.[228] Traditional Fourth Amendment analysis presumes a human actor—one who forms intent, exercises judgment, and can be held accountable for reasonableness.[229] However, as Part III illustrated, modern drones perform many of these functions through AI-Assisted Capabilities, such as facial recognition, trajectory, and object redetection.[230] When drones decide—when they initiate tracking, reacquire targets, or select camera modes based on algorithmic inference—the government has effectively delegated its surveillance authority to software.[231] This delegation raises two interrelated concerns: accountability and reasonableness.[232]
Accountability demands a discernible record of who or what made each surveillance decision.[233] Without audit logs documenting human authorization, AI activations, and overrides, it becomes impossible to trace responsibility when constitutional lines are crossed.[234] Reasonableness, in turn, depends on the foreseeability of the system’s behavior.[235] If a department deploys an autonomous function known to generate false identifications or to persistently track non-targets, that risk should inform the Fourth Amendment analysis.[236] Just as the Court in United States v. Leon conditioned good-faith reliance on objectively reasonable conduct, so too should constitutional “reasonableness” in the AI context encompass the agency’s programming choices.[237]
Delegation without procedural safeguards also implicates separation-of-powers principles.[238] The Constitution vests the warrant function in the judiciary, not in software capable of unreviewed surveillance decisions.[239] AI tools are now able to apply object recognition, re-identification, or trajectory modeling to historical footage, performing new acts of perception that were technologically impossible at the time of capture.[240] These actions are the functional equivalent of a search and the warrant process must adapt—requiring human confirmation before any AI-generated tracking or reacquisition occurs.[241] Human-in-the-loop design thus becomes not only a technical specification but a constitutional mandate.[242]
B. Persistent Aerial Archives: Cloud Retention, Retrospective Searches, and Warrants.
If autonomy transforms “who” searches, then persistence transforms “when” the search occurs.[243] The shift from real-time flight to continuous archival retention has profound constitutional consequences.[244] A drone that records every sortie to a cloud server creates a searchable time machine—a dataset that can be queried long after the original event.[245]
In Carpenter v. United States, the Court recognized that retrospective access to cell-site location data constitutes a search because it reveals “the whole of individuals’ movements.”[246] Moreover, the Fourth Circuit’s decision in Leaders of a Beautiful Struggle v. Baltimore Police Department extended that logic to aerial imagery, holding that accessing a citywide, multi-week archive was a search requiring a warrant.[247]
The case law in this area is clear, data, whether it be aerial imagery or cell-site location data, that can reveal the whole of individual movements requires a warrant.[248] Yet, Maryland’s DFR platform already stores video and coordinate metadata through Dronesense and FlightHub 2 for operational review.[249] While these functions serve legitimate accountability and training purposes, they also create the potential for retroactive investigation outside of judicial oversight.[250] Once stored, video can be queried by time, location, or object—turning operational footage into a de facto database of movements.[251] The constitutional question thus shifts from the moment of flight to the moment of retrieval.[252]
Under Carpenter and Leaders, accessing or querying an archive to reconstruct an individual’s path should be treated as a discrete search event, subject to a warrant and minimization order.[253] The warrant requirement must also extend to analytic enhancements of archived data.[254] When AI tools apply object recognition, re-identification, or trajectory modeling to historical footage, they perform new acts of perception that were technologically impossible at the time of capture.[255] A warrant authorizing a live flight does not implicitly authorize future algorithmic analysis of the resulting data.[256] Therefore, Maryland’s framework should require a separate judicial order—akin to a “secondary use warrant”—for any analytic operation performed on stored imagery.[257]
C. Heightened Protections Near the Home: Curtilage Safeguards and Thermal Limits.
Nowhere is the balance between autonomy and privacy more delicate than at the threshold of the home.[258] The Fourth Amendment and Article 26 of the Maryland Declaration of Rights both treat the home and its curtilage as zones of heightened expectation.[259] Even when said data is obtained without a physical trespass into the area of the homeowner.[260]
For instance, in Kyllo v. United States, the Court held that using a thermal imager to detect details within a home constituted a search because it revealed information “that would previously have been unknowable without physical intrusion.”[261] That reasoning applies with equal force to drones equipped with infrared or high-zoom optics capable of penetrating the home’s spatial boundary.[262] Accordingly, any drone operation targeting residential areas should be subject to automatic warrant requirements and technical minimization.[263]
Statutory rules should codify altitude ceilings, geofenced buffer zones, and automated “pause” functions whenever a flight path intersects curtilage.[264] These measures function as digital analogs to physical warrants—mechanical enforcements of constitutional restraint.[265] Drones conducting non-emergency missions should also be prohibited from activating thermal, infrared, or zoom functions near residences absent judicial approval.[266] This approach harmonizes constitutional doctrine with technical design.[267] Kyllo teaches that sense enhancement near the home demands judicial supervision; Leaders warns that persistent observation becomes a search; and Carpenter instructs that data aggregation magnifies privacy risk.[268] Together, these principles justify statutory codification of “home-adjacent minimization”—a set of automated parameters that prevent surveillance technologies from inadvertently crossing the constitutional boundary.[269]
Having articulated the constitutional imperatives of autonomy, archives, and curtilage, the analysis now moves from a judicial doctrine to legislative design.[270] Part VI advances a Maryland statute, which is modeled on Illinois’s Freedom from Drone Surveillance Act but expressly updated for artificial intelligence and supervised autonomy and that operationalizes these principles through warrant rules, analytic limits, retention schedules, and civil remedies.[271]
VI. Legislative Design: A Maryland Statute for AI‑Enabled Drone Operations.
Part VI operationalizes those imperatives.,[272] that Maryland cannot rely on case‑by‑case adjudication to regulate AI‑assisted drone use; it must legislate.[273] Fortunately, an instructive model already exists.[274] Illinois’s Freedom from Drone Surveillance Act offers a statutory framework that, if updated for artificial intelligence and supervised autonomy, can anchor Maryland’s next generation of privacy protection.[275]
- Baseline Model: Illinois’s Freedom from Drone Surveillance Act.
Illinois became the first state to enact a comprehensive statute governing police drone use.[276] The Freedom from Drone Surveillance Act embodies a warrant‑first structure with narrow, enumerated exceptions.[277] Law enforcement may deploy drones pursuant only to a warrant issued by a court on probable cause, except in circumstances of imminent danger to life, high‑risk search‑and‑rescue, disaster response, or missing‑persons investigations.[278] The Act also imposes temporal limits such as requiring that flights conducted under emergency exceptions cease once the exigency ends and mandates deletion of non‑evidentiary data within thirty days.[279]
The Illinois statute provides several key architectural lessons.[280] First, it embeds judicial authorization at the front end of drone operations, converting what might otherwise be discretionary policy into a matter of legal obligation.[281] Second, it prohibits biometric surveillance, including facial recognition and weaponization, absent specific legislative approval.[282] Third, it establishes public‑reporting duties, requiring annual disclosure of drone use, mission categories, and data retention.[283] Collectively, these provisions implement the warrant requirement ex ante: they make privacy protection a structural feature of program design, rather than a reactive judicial remedy.[284]
Additional justification for Maryland to follow Illinois lead regarding constitutional alignment is strong.[285] Both states interpret their declaration of rights to mirror the Fourth Amendment’s reasonableness test while allowing the legislature to impose stricter standards.[286] Maryland therefore stands on firm doctrinal ground to adopt Illinois’s framework as a baseline while adapting it for the unique risks of AI autonomy and cloud‑based analytics.[287]
B. Maryland Adaptation: Explicit AI Coverage, Human Confirmation, Logging, and Oversight.
A Maryland statute modeled on Illinois’s law must explicitly recognize that AI and autonomy constitute distinct regulatory categories.[288] DFR already employs algorithmic tracking and mission replay; therefore, future deployments will only deepen that integration.[289] Accordingly, Maryland’s adaptation should expand the statutory text in four dimensions:
1. Explicit AI Coverage. The law should define “autonomous or AI‑assisted drone operation” as any mission in which software initiates, continues, or modifies surveillance without contemporaneous human control.[290] This definition ensures that the statute reaches not only human‑directed flights but also automated tracking and post‑hoc analytic processing.[291]
2. Human Confirmation and Audit Logging. Any AI‑triggered surveillance action, such as automated target reacquisition or facial re‑identification, must receive affirmative human confirmation before proceeding.[292] Agencies must maintain immutable audit logs recording time stamps, operator identities, AI activations, and overrides.[293] Such logs become the evidentiary backbone for accountability and are analogous to chain‑of‑custody documentation in digital forensics.[294]
3. Oversight and Transparency. Maryland should establish a centralized oversight entity, which either could be within the Office of the State Public Defender or the Inspector General, to review drone mission data, audit retention compliance, and publish anonymized quarterly reports.[295] This ensures that transparency is not discretionary but institutionalized.[296]
4. Integration with Existing Statutes. The law should harmonize with Maryland’s Public Information Act and data‑protection statutes to ensure that citizens can access records of drone activity, subject to narrow security exceptions.[297]
Through these adaptations, Maryland’s statute would do what Illinois’s cannot: regulate not merely the use of drones but the decision logic embedded within them.[298] The Fourth Amendment’s focus on “reasonableness” thus becomes concretized in code design and administrative process.[299]
C. Enforcement: Exclusionary Rule, Civil Action, and Administrative Sanctions.
Statutory rights are hollow without remedies.[300] Illinois’s Act provides a civil enforcement mechanism and requires suppression of unlawfully obtained evidence.[301] Maryland should preserve and expand these remedies across three dimensions: first, through the exclusionary rule.[302] Courts must suppress any evidence derived from drone surveillance conducted in violation of the statute, consistent with Mapp v. Ohio’s application of the exclusionary principle to states.[303] The exclusionary rule serves not merely to deter misconduct but to maintain judicial legitimacy in the face of pervasive, technology‑driven surveillance.[304]
Secondly, it should allow for a Private Right of Action.[305] Individuals whose images, metadata, or location histories are captured without lawful authorization should have standing to seek injunctive relief and statutory damages.[306] Such a remedy extends Fourth Amendment values into the administrative state, offering a concrete path for redress even when criminal prosecution is absent.[307]
Lastly, administrative sanctions should be available.[308] Agencies that fail to maintain required logs, delete data, or obtain proper warrants should face penalties enforced by the State Ethics Commission or the Inspector General.[309] Sanctions may include budgetary withholding, license suspension for operators, and referral for disciplinary proceedings.[310] Together, these enforcement mechanisms ensure that constitutional principles are not left to post‑hoc judicial interpretation.[311] They embed privacy accountability within Maryland’s administrative fabric, transforming statutory compliance into a condition of operational legitimacy.[312]
By marrying Illinois’s statutory model to Maryland’s constitutional culture, the proposed Act moves privacy protection from abstraction to architecture.[313] Part VII extends that integration further—linking reliability, safety, and cyber‑integrity to the standard of constitutional reasonableness and ensures that drone regulation addresses not only what is seen but also how it is seen.[314]
VII. Integrating Safety, Reliability, and Reasonableness.
A statutory framework can only be as effective as the systems and institutions that implement it.[315] Part VI proposed a Maryland statute modeled on Illinois’s Freedom from Drone Surveillance Act.[316] This Part explains why the legitimacy of that framework ultimately depends on technical reliability and procedural integrity.[317] The constitutional measure of “reasonableness” under the Fourth Amendment and Article 26 of the Maryland Declaration of Rights cannot be divorced from the technological soundness of the tools the government uses.[318] When software autonomy, navigation accuracy, or cybersecurity flaws distort what the state perceives, reasonableness collapses.[319] The same principle applies institutionally: training, auditing, and transparency transform constitutional ideals into operational practice.[320]
Reliability and safety are not merely engineering concerns, they are constitutional predicates.[321] The Fourth Amendment’s reasonableness inquiry has always incorporated the manner of the search and seizure; and errors in perception, execution, or record-keeping bear directly on whether government conduct is “reasonable.”[322] For AI-enabled drones, that inquiry extends to cyber integrity, airworthiness, and compliance with federal safety regulations.[323]
Under 14 C.F.R. § 107 and the FAA’s reliability standards, every unmanned aircraft must be maintained to ensure “safe operating condition.”[324] FAA advisory circulars governing Beyond Visual Line of Sight (“BVLOS”) operations further require that any autonomous or semi-autonomous function be tested, validated, and documented prior to use.[325] These requirements are not merely bureaucratic, they delineate the technological boundary of constitutional reasonableness.[326] A surveillance program that knowingly deploys unreliable autonomy or untested object-recognition models risks unreasonable error akin to an unlawful search.[327]
Maryland law reinforces this link between technological integrity and legal responsibility.[328] Under Md. Econ. Dev. § 14-301, the State, not local governments, retains exclusive regulatory authority over the testing and operation of unmanned aircraft systems.[329] That preemption means the State must also bear responsibility for ensuring that any system it authorizes meets baseline reliability and cybersecurity standards.[330] Likewise, Md. Transp. §§ 5-1001 & 5-104 recognize a public right of flight conditioned on safety: air transit is lawful only if conducted at altitudes and in manners that are not “imminently dangerous to persons or property.”[331] These provisions effectively codify reliability, so operational safety becomes a precondition to lawful surveillance.[332] Integrating FAA and state safety frameworks into constitutional analysis ensures that “reasonableness” evolves with technology.[333] A drone that navigates erratically, loses telemetry, or transmits unsecured video cannot form the factual foundation of a constitutionally valid search.[334]
Reliability is as much institutional as technical.[335] Just as United States v. Leon conditioned good-faith reliance on objectively reasonable conduct, constitutional reasonableness in the age of autonomy requires reasonable administration.[336] Maryland can institutionalize reliability through three mutually reinforcing mechanisms: operator training, AI validation, and periodic audits.[337]
The first is operator training.[338] The Md. Public Safety § 3-506 precedent for mandatory eyewitness-identification policies provides an instructive analogue.[339] Just as every law-enforcement agency must adopt Department of Justice–compliant eyewitness procedures to ensure accuracy, every agency employing drones should be statutorily required to adopt standardized AI-use protocols approved by the Maryland Police Training and Standards Commission.[340] Certification should include both technical proficiency and legal comprehension-ensuring that operators understand how warrant rules, curtila ge protections, and minimization requirements apply to AI-enabled sensors.[341]
The second is AI validation.[342] Before deployment, all machine-learning models used for object detection, person re-identification, or trajectory prediction should undergo documented validation testing akin to evidentiary reliability under Daubert v. Merrell Dow Pharmaceuticals, Inc.[343] Validation results should be retained as part of the system’s audit file and made available for judicial review when AI outputs form part of probable-cause or evidentiary determinations, similar to breath analyzers, or Radar speed guns.[344]
The third mechanism is the use of reliability audits.[345] Annual third-party audits should review each agency’s compliance with FAA maintenance requirements, cybersecurity controls, and AI-model updates.[346] These findings should be transmitted to the oversight body established under the proposed statute.[347] Regular audits ensure that reliability is not a one-time certification but a continuing constitutional practice.[348]
Transparency converts reliability into legitimacy.[349] Constitutional reasonableness depends not only on internal accuracy but also on public trust.[350] Maryland’s Transportation statutes already acknowledge a public right of air transit; that right should be paired with a public right of information.[351] Agencies operating drones should publish annual reports detailing flight counts, mission types, warrant authorizations, retention statistics, and any AI-model changes.[352] Civilian oversight bodies—whether nested within the Inspector General or constituted as an independent Privacy and Technology Commission—should possess subpoena power to inspect logs, audit compliance, and investigate complaints.[353] Public summaries of such audits, anonymized for security, create the transparency that sustains democratic legitimacy.[354] Finally, transparency must extend to intergovernmental communication.[355] Federal preemption and state sovereignty coexist uneasily under Md. Transp. § 5-104, which vests airspace sovereignty in the State subject to federal delegation.[356] Regular coordination with the FAA, combined with public reporting, ensures that Maryland’s regulatory scheme remains consistent with federal standards while responsive to local privacy expectations.[357] In this sense, transparency is not ancillary to reasonableness, rather it is reasonableness, expressed institutionally.[358]
As AI autonomy and cloud analytics redefine the scope of surveillance, safety and reliability become the new grammar of constitutional reasonableness.[359] The proposed Maryland statute cannot succeed unless its operational ecosystem of technical, institutional, and public, is equally sound.[360] The final Part concludes by situating Maryland’s model within the broader national movement toward anticipatory governance, where the rule of law adapts at the speed of the machine.[361]
VIII. Conclusion: Designing Constitutional Reasonableness for the Machine Age.
Maryland’s moment mirrors the very arc that opened this Article.[362] In the span of two days, Air Force officials and President Zelenskyy warned that the speed of AI now outstrips the speed of human judgment—and that autonomy, once embedded in sensing and targeting systems, reshapes the moral and legal premises on which decision making occurs.[363] What begins on foreign battlefields rarely stays there.[364] The same classes of perception, prediction, and task-level autonomy that now drive military planning cycles migrate quickly into domestic public-safety tools.[365] Montgomery County’s DFR program already illustrates this dynamic: rapid launch, automated stabilization, intelligent tracking, and cloud-based video create operational value, but also the architecture for persistent monitoring if left unguided.[366]
The constitutional doctrines canvassed throughout this Article, including Ciraolo and Riley’s tolerance for brief aerial glances, Kyllo’s protection for the home and curtilage, Jones and Carpenter’s concern with long-term technological tracking, and the Fourth Circuit’s recognition in Leaders of a Beautiful Struggle that access to an aerial archive is itself a search, were never designed for autonomous systems capable of sustaining continuous presence, generating probabilistic inferences, or delegating targeting decisions to software.[367] The doctrinal guardrails exist, but they were built for pilots, not programs, and for officers whose observations are episodic, but not for drones whose senses and memory never fatigue.[368]
As this Article has argued, the doctrinal mismatch creates a narrow window in which Maryland can act.[369] The DFR program is still young; its technical stacks still malleable, and its institutional norms are still forming.[370] This is precisely the moment when statutory architecture can shape technological evolution rather than merely respond to it.[371] A comprehensive drone statute that is adapted from Illinois’ but expressly tailored to AI-assisted and autonomy-capable platforms, would do more than codify warrant requirements or retention limits.[372] It would constitutionalize system design;[373] embed minimization into software parameters; require audit logs for every analytic query; prohibit person-reidentification and facial recognition without judicial authorization; and tether automated tracking to human-in-the-loop confirmation.[374] These are not after-the-fact remedies.[375] They are the engineering-level equivalents of the Fourth Amendment’s core commitments.[376]
Designing reasonableness for autonomous systems therefore involves more than constraining police conduct.[377] It requires specifying how systems perceive, what they remember, and when they may act.[378] It requires aligning software defaults with constitutional values before scale, automation, and institutional inertia make such alignment impossible.[379] Additionally, it requires public transparency through dashboards, retention disclosures, and audit reports so affected communities can see, not only what drones captured, but how and why.[380]
If Maryland adopts such a framework, it will not simply harmonize drone operations with existing federal doctrine.[381] It will lead.[382] The state can model a form of anticipatory governance that recognizes both the promise of AI-enabled response and the peril of unbounded, machine-driven surveillance.[383] It can demonstrate that innovation and constitutional fidelity are not opposing forces but mutually reinforcing conditions for public trust.[384] Furthermore, it can show that in the machine age, legitimacy derives not only from lawful outcomes but also from lawful architectures—systems whose constraints are as transparent as their capabilities.[385]
The future of public safety will increasingly involve software that perceives, recommends, and, eventually, acts.[386] The question is whether that software will be shaped by constitutional principles or merely judged against them after the fact. By acting now, before autonomy becomes entrenched and before persistent archives become the default, Maryland can ensure that when drones begin to decide, they do so within a constitutional frame that the legislature, the courts, and the public helped build.[387] In doing so, Maryland will not only comply with the Fourth Amendment but it will also help define the state’s next chapter.[388]
Attachment A: Model Maryland AI-Enabled Drone Surveillance Statute
Section 1. General Warrant Requirement.
(a) Except as provided in Section 2, a law enforcement agency may not operate an unmanned aircraft system for surveillance, investigation, evidence gathering, or data collection unless the agency has obtained a valid warrant based on probable cause.
Section 2. Exceptions to Warrant Requirement.
(a) A law enforcement agency may operate an unmanned aircraft system without a warrant only under the following circumstances:
- Exigent Circumstances. When a law enforcement officer reasonably believes that immediate deployment is necessary to:
- Prevent imminent danger of death or serious bodily injury to any person;
- Forestall the imminent escape of a suspect; or
- Prevent the imminent destruction of evidence of a felony offense.
- Search and Rescue. To locate a missing person, provided that:
- There is reasonable suspicion that the person’s safety is at risk;
- The deployment is limited in duration and geographic scope to the area where the person is reasonably believed to be located; and
- Any footage or data collected incidentally to the search is subject to the retention and deletion requirements.
- Disaster and Emergency Management. To assess damage, coordinate emergency response, or protect public safety during a natural disaster, fire, hazardous materials incident, or other emergency officially declared by a competent authority.
- Crime Scene Documentation. To photograph or document a crime scene, collision scene, or other site requiring forensic preservation, provided that:
- The deployment occurs only after law enforcement has lawfully secured the scene;
- The unmanned aircraft does not surveil persons or areas beyond the secured perimeter; and
- Sense-enhancement technologies are not activated outside the lawful perimeter.
- Drone As First Responders. To photograph or document a crime scene, collision scene, or other site requiring forensic preservation, provided that:
- The deployment occurs only after law enforcement has lawfully secured the scene;
- The unmanned aircraft does not surveil persons or areas beyond the secured perimeter; and
- Sense-enhancement technologies are not activated outside the lawful perimeter.
(b) For deployments under subsection (a)(1) (exigent circumstances), the law enforcement agency must:
- Document specific facts constituting the exigency in a written report filed within 48 hours of the deployment;
- Apply for a warrant as soon as practicable if continued surveillance is necessary; and
- Terminate the operation immediately upon cessation of the exigent circumstances.
Section 3. Restrictions on Artificial Intelligence, Biometric, and Sense-Enhancement Technologies.
(a) Prohibition on Facial Recognition and Person Re-Identification. A law enforcement agency may not:
- Equip an unmanned aircraft system with facial recognition technology or biometric identification technology; or
- Use person re-identification algorithms or predictive analytics to identify, track or correlate individuals across multiple unmanned aircraft deployments or archived data, unless the agency obtains a court order authorizing such use pursuant to Subsection 3.
(b) Court Authorization for Biometric Technologies.
A law enforcement agency seeking to use biometric identification technology or person re-identification analytics in connection with unmanned aircraft operations must obtain a court order based on probable cause.
(c) Prohibition on Sense-Enhancement Technologies NearsHomes and Curtilage.
- A law enforcement agency may not activate thermal imaging, infrared sensors, or other sense-enhancement technology to observe the interior of dwelling or its curtilage unless:
- The agency has obtained a warrant specifically authorizing use of sense-enhancement technology; or
- Exigent circumstances exist as defined in Section 2.
(d) Any order issued under this section must:
- Specify the duration of authorization, which may not exceed 30 days without judicial review;
- Require the law enforcement agency to file periodic compliance reports; and
- Include conditions ensuring human verification of all algorithmic identification before investigative or enforcement action is taken.
Section 4. Human-in-the-Loop Requirements for Autonomous Operations.
(a) A law enforcement agency employing autonomous or AI-assisted unmanned aircraft operations must ensure that:
- Any surveillance, tracking or targeting action initiated by artificial intelligence receives affirmative human confirmation before execution;
- A certified remote pilot in command retains real time-time override capability throughout the operation; and
- ——
Section 5. General Retention, Deletion, and Aerial Archive Requirements.
(a) Non-evidentiary data must be automatically deleted within 30 days.
(b) Querying or analyzing stored UAS footage or metadata to reconstruct a person’s movements constitutes a search and requires a warrant.
(c) AI-based analysis of archived data requires:
(1) A warrant; and
(2) Compliance with Section 5.
(d) Data collected under exigent or emergency exceptions may not be used for unrelated investigations.
Section 6. Civil Action for Damages.
(a) An individual whose image, biometric data, location information, or coordinate metadata is collected, retained, analyzed or disseminated in violation of the statute may bring a civil action in a court of competent jurisdiction against
- The law enforcement agency responsible for the violation; and
- Any officer, employee, or agent who knowingly, recklessly, or negligently caused or contributed to the violation.
(b) A plaintiff who prevails in an action under this section is entitled to:
Section 7. Security, Cyber-Integrity, and System Reliability Requirements.
(a) All UAS systems must comply with FAA reliability and maintenance standards.
(b) Agencies must implement:
(1) Encryption of all transmitted and stored UAS data;
(2) Access-control logs; and
(3) Annual third-party audits.
(c) Evidence obtained through unreliable or unvalidated AI models is inadmissible.
Section 8. Transparency, Public Reporting, and Oversight.
(a) Each agency shall publish quarterly public reports including:
(1) Number of deployments;
(2) Mission categories;
(3) Warrants obtained;
(4) Use of analytic or biometric technologies; and
(5) Retention and deletion statistics.
(b) The Maryland Inspector General for Public Safety shall:
(1) Conduct annual audits;
(2) Possess subpoena authority; and
(3) Publish an annual public report.
Section 9. Enforcement; Remedies; Exclusionary Rule.
(a) Evidence obtained in violation of this Act, and any derivative evidence, shall be suppressed.
(b) Any individual whose data is unlawfully collected may bring a civil action against:
(1) The responsible agency; and
(2) Any officer who knowingly, recklessly, or negligently caused the violation.
(c) Remedies include:
(1) Actual damages or statutory damages of $1,000 per violation;
(2) Punitive damages for willful violations;
(3) Injunctive or declaratory relief; and
(4) Attorney’s fees and costs.
(d) Agencies may face budgetary penalties or operator-license suspension.
Section 12. Severability.
(a) If any provision of this Act is held invalid, the remaining provisions remain in force.
Section 13. Definitions.
(a) “Unmanned Aircraft System” (UAS) means an aircraft operated without the possibility of direct human intervention from within or on the aircraft, including any associated elements such as communication links, onboard sensors, data storage, and remote-control equipment.
(b) “AI-Assisted Operation” means any UAS function in which software performs perception, tracking, reacquisition, routing, or analytic tasks that materially influence surveillance or data collection.
(c) “Autonomous Operation” means any UAS function that initiates, continues, or modifies surveillance, tracking, or data capture without contemporaneous human command.
(d) “Biometric Identification Technology” means facial recognition, person re-identification, gait recognition, or any algorithmic system that identifies or attempts to identify a specific individual.
(e) “Curtilage” has the same meaning as recognized under Article 26 of the Maryland Declaration of Rights.
(f) “Sense-Enhancement Technology” includes thermal imaging, infrared sensors, telescopic or ultra-zoom optics, or other devices not in general public use that can reveal details of a dwelling or its curtilage otherwise unknowable without physical intrusion.
(g) “Surveillance” means the intentional collection, recording, or retention of images, audio, coordinate metadata, or other data relating to persons, structures, or vehicles.
Section 14. Effective Date.
(a) This Act shall take effect October 1, 2026.
*Professor Simon M. Caine, Esquire is an Associate Counsel in the Office of the General Counsel for the Department of the Navy, where he litigates complex federal employment matters and other civil litigation. Mr. Caine is an adjunct professor of law and has taught legal writing, advocacy, and professional skills at the Catholic University of America, Columbus School of Law, and the University of Baltimore School of Law. His teaching emphasizes statutory interpretation, constitutional litigation, and disciplined legal writing, and he regularly mentors law students and junior attorneys in scholarly research and appellate practice. In addition to his civilian practice and teaching, Mr. Caine serves as a Major in the United States Air Force Reserve Judge Advocate General’s Corps. He is a certified visual flight rules (“VFR”) pilot and a certified unmanned aircraft systems (“UAS”) pilot, with a sustained interest in aviation and emerging technologies. His scholarship focuses on the intersection of artificial intelligence, aerial surveillance, and constitutional criminal procedure, with particular attention to state and local regulatory design and the implications of rapid technological change for Fourth Amendment doctrine. Mr. Caine wishes to thank his research assistants, Chloe Curtis and Annie Segall, for their contributions to this article. The views expressed in this Article are solely those of the author and do not necessarily represent the views of the Department of War, the Department of the Navy, or the United States Government. The author writes in a personal capacity.
[1] See discussion infra notes 4–7, 13–15.
[2] See discussion infra notes 4-10.
[3] AI hallucination is a phenomenon where, in a large language model (“LLM”) often a generative AI chatbot or computer vision tool, perceives patterns or objects that are nonexistent or imperceptible to human observers, creating outputs that are nonsensical or altogether inaccurate. What Are AI Hallucinations?, IBM, https://www.ibm.com/think/topics/ai-hallucinations (on file with the University of Baltimore Law Forum) (last visited Apr. 14, 2026).
[4] Shaun Waterman, Air Force Grapples with Limitations of AI, Air & Space Forces Mag. (Sept. 23, 2025), https://www.airandspaceforces.com/air-force-grapples with-limitations-of-ai/ (on file with the University of Baltimore Law Forum).
[5] See id.
[6] Id. (“At a recent experiment staged by the Advanced Battle Management System Cross-Functional Team, vendors and Air Force coders used AI to do the work of [battle managers]—deciding which platforms and weapons systems should be used against a particular target and generating Courses of Action, or COAs, to achieve a military objective.”) (citing Greg Hadley, What Does the Future Look Like for Battle Managers?, Air & Space Forces Mag. (Feb. 16, 2024), https://www.airandspaceforces.com/usaf-c2-battle-managers-future/ (on file with the University of Baltimore Law Forum)).
[7] Id.
[8] Alex Leff, Ukraine’s Zelenskyy issues a stark warning about a global arms race and AI war, NPR (Sept. 24, 2025, at 11:41 ET), https://www.npr.org/2025/09/24/nx-s1-5552206/ukraine-zelenskyy-un-general-assembly-trump (on file with the University of Baltimore Law Forum).
[9] Id.
[10] See Morris Tidball-Binz (Special Rapporteur on Extrajudicial, Summary or Arbitrary Executions), Autonomous Weapons Systems, at 7, 9, U.N. Doc. A/HRC/56/CRP.5 (Apr. 16, 2024).
[11] See infra notes 12-19.
[12] Montgomery County Police Drone Program Marks One Year Milestone, Montgomery Cnty. Police Dep’t: News Releases (Nov. 20, 2024), https://www2.montgomerycountymd.gov/mcgportalapps/Press_Detail_Pol.aspx?Item_ID=46189 (on file with the University of Baltimore Law Forum) [hereinafter MCPD Press Release]; see Drone as First Responder – Flight Maps & Data Page, Montgomery Cnty. Police Dep’t, https://dashboard.dronesense.com/mcpddfr (on file with the University of Baltimore Law Forum) (last visited Nov. 12, 2025) [hereinafter DFR Public Dashboard].
[13] DFR Public Dashboard, supra note 12 (showing a total of 198 flights for September 2025).
[14] MCPD Press Release, supra note 12.
[15] See id.
[16] See discussion infra notes 17-19.
[17] See, e.g., California v. Ciraolo, 476 U.S. 207, 215 (1986); Florida v. Riley, 488 U.S. 445, 455 (1989); Dow Chem. Co. v. United States, 476 U.S. 227, 232 (1986); Kyllo v. United States, 533 U.S. 27, 40 (2001); United States v. Knotts, 460 U.S. 276, 284-85 (1983); United States v. Karo, 468 U.S. 705, 721 (1984); United States v. Jones, 565 U.S. 400, 412-13 (2012); Carpenter v. United States, 585 U.S. 296, 320 (2018).
[18] See cases cited supra note 17.
[19] See cases cited supra note 17.
[20] Leaders of a Beautiful Struggle v. Balt. Police Dep’t, 2 F.4th 330, 333 (4th Cir. 2021) (en banc).
[21] See infra Parts II-IV.
[22] See infra Parts V-VI.
[23] See infra Part VI; Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/10, /15 (West 2025).
[24] See infra Part VI.
[25] See supra Abstract.
[26] See infra Part II.
[27] See infra Part II.
[28] See infra Part III.
[29] See infra Part IV.
[30] See infra Part V.
[31] See infra Part VI.
[32] See infra Part VII.
[33] See infra Part VIII.
[34] See discussion infra notes 35-41.
[35] See infra Parts II–III.
[36] See infra Part II.
[37] See infra Part IV.
[38] See infra Part III.
[39] See infra Part VI.
[40] See infra Part VI; Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/10, /15 (West 2025).
[41] See infra Part VI.
[42] See infra Part II.
[43] See infra Part II.
[44] See infra Part II.
[45] See discussion infra notes 46-50.
[46] See MCPD Press Release, supra note 12; DFR Public Dashboard, supra note 12.
[47] MCPD Press Release, supra note 12; Drone as First Responder (DFR) Program, Montgomery Cnty. Police Dep’t, https://www.montgomerycountymd.gov/pol/howdoI/drone-as-first-responder.html (on file with the University of Baltimore Law Forum) (last visited Nov. 15, 2025) [hereinafter MCPD DFR Program]; DJI’s New Matrice 350 RTK Redefines the Ultimate Tool for the Job, DJI Enter. (May 18, 2023), https://enterprise.dji.com/news/detail/matrice-350-rtk (on file with the University of Baltimore Law Forum) [hereinafter DJI, Latest Update on the Matrice 350 RTK].
[48] DFR Public Dashboard, supra note 12.
[49] See MCPD Press Release, supra note 12; MCPD DFR Program, supra note 47; DJI, Latest Update on the Matrice 350 RTK, supra note 47.
[50] See Henry White, DFR in action: Inside Montgomery County PD’s Drone as First Responder program, Police1 (Feb. 25, 2025, at 16:28 ET), https://www.police1.com/drones/dfr-in-action-inside-montgomery-county-pds-drone-as-first-responder-program (on file with the University of Baltimore Law Forum).
[51] See infra Part III.
[52] See infra Parts III-IV.
[53] See discussion infra notes 54-56; infra Part IV.
[54] California v. Ciraolo, 476 U.S. 207, 213-14 (1986); Florida v. Riley, 488 U.S. 445, 449-50 (1989); Dow Chem. Co. v. United States, 476 U.S. 227, 236 (1986).
[55] See Ciraolo, 476 U.S. at 209 (showing aerial surveillance using private airplane at 1,000 feet altitude for naked-eye observation and photography of defendant’s backyard); Riley, 488 U.S. at 451 (showing police helicopter surveillance at 400 feet altitude using naked-eye observation through greenhouse roof openings); Dow Chem., 476 U.S. at 238 (showing EPA aerial surveillance of industrial complex using standard floor-mounted precision aerial mapping camera from altitudes of 12,000, 3,000, and 1,200 feet).
[56] Compare sources cited supra note 55 (technology with human oversight), with MCPD DFR Program, supra note 47 (automatic AI-assisted drones).
[57] United States v. Jones, 565 U.S. 400, 403 (2012); Carpenter v. United States, 585 U.S. 296, 310 (2018).
[58] Jones, 565 U.S. at 403; Carpenter, 585 U.S. at 310.
[59] See, e.g., Jones, 565 U.S. at 411-13; Carpenter, 585 U.S. at 309.
[60] Leaders of a Beautiful Struggle v. Balt. Police Dep’t, 2 F.4th 330, 345–52 (4th Cir. 2021) (en banc).
[61] See id.
[62] See id.; infra notes 63-64.
[63] See MCPD DFR Program, supra note 47.
[65] See Police Drones Are Flying Over Maryland. Will State Lawmakers Regulate Them?, CityBiz (Jan. 24, 2025), https://www.citybiz.co/article/652346/police-drones-are-flying-over-maryland-will-state-lawmakers-regulate-them/ (on file with the University of Baltimore Law Forum); infra Part VI.
[66] See infra notes 67-70.
[67] Marvin Hanisch et al., Digital governance: A conceptual framework and research agenda, J. of Bus. Rsch., at § 3.2 (Jul. 2023) (first citing Alexander Benlian et al., Algorithmic Management: Bright and Dark Sides, Practical Implications, and Research Opportunities, 64 Bus. & Info. Sys. Eng’g 825 (2022); and then citing Ronan Renwick & Robert Gleasure, Those Who Control the Code Control the Rules: How Different Perspectives of Privacy Are Being Written into the Code of Blockchain Systems, 36 J. of Info. Tech. 16 (2021)).
[68] Maneka Sinha, The Automated Fourth Amendment, 73 Emory L. J. 589, 630 (2024); see infra Part V.
[69] Sinha, supra note 66, at 630; see infra Part V.
[70] See discussion infra notes 71-73.
[71] See infra Parts III-V.
[72] See infra Parts V-VI.
[73] Leaders of a Beautiful Struggle v. Balt. Police Dep’t, 2 F.4th 330, 333 (4th Cir. 2021) (en banc).
[74] See infra Part III.
[75] See infra Parts V-VI.
[76] See infra Part VI. See generally Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/10, /15 (West 2025) (prohibiting the use of drones except as provided in Section 15 of the statute).
[77] See infra Part III.
[78] See infra Part III.
[79] See infra Part III.
[80] See infra Part III.
[81] See infra Part III.
[82] See infra Part III.
[83] See MCPD DFR Program, supra note 47; supra Part II.
[84] See infra Part III.
[85] See infra Sections III.A-C.
[86] Artificial Intelligence Risk Management Framework (AI RMF 1.0), Nat’l Inst. of Standards & Tech. 40 (U.S. Dep’t of Com. Jan. 2023), https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf (on file with the University of Baltimore Law Forum) [hereinafter NIST, Artificial Intelligence Risk Management Framework].
[87] U.S. Dep’t of Def., DOD Directive 3000.09, Autonomy in Weapon Sys. (2023).
[88] MCPD DFR Program, supra note 47.
[89] FlightHub 2 Overview, One-Stop Aircraft Task Management Cloud Platform, DJI Enter., https://enterprise.dji.com/flighthub-2 (on file with the University of Baltimore Law Forum) (last visited Nov. 20, 2025) [hereinafter FlightHub 2 Platform].
[90] See id.
[91] See 14 C.F.R. § 107.19 (2021).
[92] See NIST, Artificial Intelligence Risk Management Framework, supra note 86, at 40.
[93] See id. at 4.
[94] See id. at 39.
[95] See supra Section III.A.
[96] Lee Lerussi, Drone forensics: How police use digital evidence from drones to fight crime and terrorism, Police1 (Aug. 23, 2025, at 17:15 ET), https://www.police1.com/investigations/drone-forensics-how-police-use-digital-evidence-from-drones-to-fight-crime-and-terrorism (on file with the University of Baltimore Law Forum).
[97] See Teaganne Finn & Ian Smalley, What is predictive analytics?, IBM, https://www.ibm.com/think/topics/predictive-analytics (on file with the University of Baltimore Law Forum) (last visited Apr. 8, 2026).
[98] Mobile SDK V5, DJI Dev., https://developer.dji.com/mobile-sdk/ (on file with the University of Baltimore Law Forum) (last visited Nov. 20, 2025); Cloud API, DJI Dev., https://developer.dji.com/cloud-api/ (on file with the University of Baltimore Law Forum) (last visited Nov. 20, 2025).
[99] NIST, Artificial Intelligence Risk Management Framework, supra note 86, at 14.
[100] U.S. Dep’t of Com., Nat’l Inst. of Standards & Tech., Face Recognition Vendor Test (FRVT), Part 3: Demographic Effects (2019) (on file with the University of Baltimore Law Forum) [hereinafter NIST, FRVT Part 3].
[101] Id. at 7.
[102] Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/10, /20, /35 (West 2025).
[103] See discussion infra notes 106-18.
[104] See Josh Rosenthal, Police drone helps convict man in Silver Spring stabbing, FOX 5 DC (Nov. 1, 2024, at 22:15 ET), https://www.fox5dc.com/news/police-drone-helps-convict-man-silver-spring-stabbing (on file with the University of Baltimore Law Forum); see also Mike Hellgren, 2 juveniles arrested in Harford County as more police departments use drones to fight crime, CBS Balt. (Nov. 28, 2024, at 18:28 ET), https://www.cbsnews.com/baltimore/news/maryland-harford-county-drone-arrest-juvenile-crime-prevention/ (on file with the University of Baltimore Law Forum).
[105] Rosenthal, supra note 104; Hellgren, supra note 104.
[106] See discussion infra notes 107-18.
[107] Rosenthal, supra note 104.
[108] Id.
[109] Id.
[110] See id.
[111] Hellgren, supra note 104.
[112] Id.
[113] Id.
[114] Id.
[115] Id.
[116] See sources cited supra notes 104.
[117] Rosenthal, supra note 104.
[118] See discussion supra notes 106-17; supra Section III.B.
[119] Face Recognition Technology Evaluation (FRTE) 1:1 Verification, Nat’l Inst. of Standards & Tech., https://pages.nist.gov/frvt/html/frvt11.html (on file with the University of Baltimore) (last visited Nov. 20, 2025) [hereinafter NIST, FRTE 1:1 Verification].
[120] NIST, FRVT Part 3, supra note 100, at 2–4; NIST, FRTE 1:1 Verification, supra note 119.
[121] NIST, FRTE 1:1 Verification, supra note 119.
[122] Id.
[123] See Leaders of a Beautiful Struggle v. Baltimore Police Dep’t, 2 F.4th 330, 347 (4th Cir. 2021) (en banc); NIST, FRVT Part 3, supra note 100.
[124] See Comment, Fourth Circuit Holds Warrantless Access of Aerial Surveillance Data Unconstitutional: Leaders of a Beautiful Struggle v. Baltimore Police Department, 135 Harv. L. Rev. 920, 923 (2022) [hereinafter Fourth Circuit Holds Warrantless].
[125] See Drones as First Responders Act, Pub. Act 103-0101 § 5, 2023 Ill. Laws (codified as amended at 725 Ill. Comp. Stat. 167/15 (West 2024)) (prohibiting law enforcement from equipping drones with weapons or facial recognition technology, subject to narrow exceptions).
[126] See id.
[127] See discussion infra notes 128-32.
[128] Criminal Procedure – Facial Recognition Technology – Requirements, Procedures, and Prohibitions, ch. 808, 2024 Md. Laws 1 (codified at Md. Code Ann., Crim. Proc. §§ 2-501-510 (West 2024)) (emphasizing § 2-504, which imposes disclosure duties in criminal and juvenile proceedings).
[129] See Crim Proc. § 2-510.
[130] See generally id. (requiring law enforcement agencies using facial recognition technology to publish annual reports disclosing certain information).
[131] See Insun Park, Beyond the Face: Examining Legal Regulations for Law Enforcement Facial Recognition Technology, 58 Akron L. Rev. 251, 277–78 (2025).
[132] See, e.g., Fourth Circuit Holds Warrantless, supra note 124, at 920-21 (describing Baltimore’s aerial program as capable of tracking movements of virtually all city residents).
[133] See supra Sections III.A–C.
[134] See supra Part III.
[135] Carpenter v. United States, 585 U.S. 296, 309-15 (2018); Leaders of a Beautiful Struggle v. Baltimore Police Dep’t, 2 F.4th 330, 338–40 (4th Cir. 2021) (en banc).
[136] Kyllo v. United States, 533 U.S. 27, 34–40 (2001).
[137] See, e.g., Cong. Rsch. Serv., R47660, Law Enforcement and Technology: Use of Unmanned Aircraft Systems (UAS), at 4, 7–9 (2023) (explaining that FAA regulation principally addresses airspace safety, leaving privacy regulation largely to states and localities).
[138] Id.
[139] See supra Section III.D.
[140] See supra Part III.
[141] See infra Part IV.
[142] See California v. Ciraolo, 476 U.S. 207, 211-12, 221-23 (1986); Florida v. Riley, 488 U.S. 445, 455 (1989); Dow Chem. Co. v. United States, 476 U.S. 227, 235–36 (1986); United States v. Jones, 565 U.S. 400, 404-05 (3023); Carpenter v. United States, 585 U.S. 296, 300-03 (2018); Leaders of a Beautiful Struggle v. Balt. Police Dep’t, 2 F.4th 330, 339-41 (4th Cir. 2021) (en banc).
[143] See infra Part IV.
[144] See infra Part IV.
[145] See infra Part IV.
[146] See discussion and cases cited infra notes 147–51.
[147] Ciraolo, 476 U.S. at 211–12, 221–23.
[148] Id. at 213–14.
[149] Florida v. Riley, 488 U.S. 445, 451 (1989).
[150] Dow Chem. Co. v. United States, 476 U.S. 227, 239 (1986).
[151] See discussion and cases cited supra notes 147-50.
[152] See discussion supra notes 153-56.
[153] See Kyllo v. United States, 533 U.S. 27, 34–40 (2001); Carpenter v. United States, 585 U.S. 296, 309-15 (2018); Leaders of a Beautiful Struggle v. Baltimore Police Dep’t, 2 F.4th 330, 338–40 (4th Cir. 2021) (en banc).
[154] See cases cited supra notes 130-33 (explaining various technology used previously before 2013, when new more affordable and accessible technology became available ); The Complete History of Drone Technology, Zenatech, https://www.zenatech.com/history-of-drone-technology/ (on file with the University of Baltimore) (last visited Dec. 7, 2024) (explaining the proliferation of affordable, camera-equipped quadcopters beginning with Parrot AR Drone in 2010 followed by DJI Phantom 1 in 2013, marking widespread consumer drone adoption).
[155] See supra Part IV.
[156] Compare cases cited supra notes 135-37 (explaining various technology used previously before 2013, when new more affordable and accessible technology became available), with MCPD DFR Program, supra note 47 (noting that the entity may have the future capability for continuous surveillance).
[157] See infra Section IV.B.
[158] Kyllo, 533 U.S. at 40.
[159] Id. at 29.
[160] Id.
[161] Id.
[162] Id.
[163] Id. at 30.
[164] Kyllo, 533 U.S. at 30.
[165] Id. at 34.
[166] Id. at 37.
[167] Id. at 40.
[168] Id. at 34.
[169] Id. at 35-36.
[170] See Kyllo, 533 U.S. at 40; Jardines v. State, 73 So. 3d 34, 43 (2011); United States v. Hay, 95 F.4th 1304, 1317 (2024).
[171] Kyllo, 533 U.S. at 37.
[172] Id.
[173] Id. at 35-36.
[174] See id. at 40; Jardines, 73 So. 3d at 43; Hay, 95 F.4th at 1317.
[175] Kyllo, 533 U.S. at 37-38.
[176] Id. at 34.
[177] Id. at 36.
[178] See discussion infra notes 179-82.
[179] See Kyllo, 533 U.S. at 40 (explaining implications of using an object not in the general public use for surveillance).
[180] Id. at 35-37.
[181] Id. at 37-38.
[182] Md. Const. Decl. of Rts., art. 26.; see also Richardson v. State, 274 A.3d 1143, 1149 (Md. 2022) (explaining the Fourth Amendment analysis).
[183] See infra Section IV.C.
[184] See infra Section IV.C.
[185] United States v. Knotts, 460 U.S. 276, 285 (1983).
[186] Id. at 281.
[187] Id. at 282–83.
[188] Id. at 283–84.
[189] United States v. Karo, 468 U.S. 705, 734-35 (1984).
[190] Id. at 714–15.
[191] Id. at 712–13, 715.
[192] Id. at 714–16.
[193] Id. at 715.
[194] United States v. Jones, 565 U.S. 400, 402 (2012).
[195] Id. at 412-13.
[196] Id. at 404–06.
[197] Id. at 408–09.
[198] Id. at 415–18 (Sotomayor, J., concurring); id. at 430–31 (Alito, J., concurring in the judgment).
[199] See Carpenter v. United States, 585 U.S. 296, 296-97 (2018).
[200] Id. at 311, 315.
[201] Id. at 320.
[202] See Leaders of a Beautiful Struggle v. Balt. Police Dep’t, 2 F.4th 330, 342 (4th Cir. 2021) (en banc).
[203] See supra Sections IV.A-C.
[204] See infra Section IV.D.
[205] See supra Sections IV.A-C.
[206] See infra Section IV.D.
[207] Leaders of a Beautiful Struggle, 2 F.4th at 342.
[208] Id. at 334, 340-42.
[209] Id. at 342–43 (first citing Carpenter v. United States, 585 U.S. 296, 310-12 (2018); and then citing United States v. Jones, 565 U.S. 400, 429 (2012)).
[210] See Interview with Capt. Victor A. Galladora, Montgomery Cnty. Police Dep’t (Oct. 31, 2025) (conducted via Microsoft Teams) (notes on file with author).
[211] Id.
[212] See California v. Ciraolo, 476 U.S. 207, 213 (1986); Florida v. Riley, 488 U.S. 445, 451 (1989); see also Leaders of a Beautiful Struggle, 2 F.4th at 342 (distinguishing AIR’s broad coverage from these precedents).
[213] Leaders of a Beautiful Struggle, 2 F.4th at 333.
[214] See Interview with Capt. Victor A. Galladora, supra note 210, at Sections VI-VII (detailing internal restrictions on zoom and thermal imaging near residences, training on constitutional limits, and privacy policies); infra Part V.
[215] See supra Part IV; infra Part V.
[216] See infra Parts V-VI.
[217] See supra Part IV.
[218] See supra Part IV.
[219] Orwin S. Kerr, The Fourth Amendment and New Technologies: Constitutional Myths and the Case for Caution, 102 Mich. L. Rev. 801, 868 (2004).
[220] Latin for “before the event.” See Ex ante, LexisNexis, https://www.lexisnexis.co.uk/legal/glossary/ex-ante (on file with the University of Baltimore Law Forum) (last visited on Apr. 15, 2026). This refers to the prospective establishment of rules, as opposed to ex post (“after the event”) judicial review that analyzes conduct after it has occurred. Ex post facto, Justia, https://dictionary.justia.com/ex-post-facto (on file with the University of Baltimore Law Forum) (last visited on Apr. 15, 2026).
[221] Kerr, supra note 219, at 868.
[222] Id. at 870.
[223] See supra Part IV; infra Part V.
[224] See infra Parts V-IV.
[225] See infra Part V.
[226] See infra Part V.
[227] See infra Part V.
[228] See infra Section V.A.
[229] See supra Sections IV.A-C.
[230] See supra Part III.
[231] See supra Section IV.D; infra Section V.A.
[232] See discussion infra notes 233-37.
[233] Madalina Busuioc, Accountable Artificial Intelligence: Holding Algorithms to Account, 81 Pub. Admin. Rev. 825, 828 (2021).
[234] See infra Section VI.B.2.
[235] Foreseeability: Explained, Cʟᴇᴀʀ Lᴇɢᴀʟ (Oct 28, 2024), https://clearlegal.com/glossary/foreseeability/ (on file with the University of Baltimore Law Forum).
[236] Kelsey Y. Santamaria, Facial Recognition Technology and Law Enforcement: Select Constitutional Considerations, Libr. of Cong. (Sept. 24, 2020), https://www.congress.gov/crs-product/R46541 (on file with the University of Baltimore Law Forum).
[237] See United States v. Leon, 468 U.S. 897, 922 (1984).
[238] See infra Section V.A; see also U.S. Const. amend. IV (requiring probable cause to be founded upon oath or affirmation, which is granted through a detached judge).
[239] See generally U.S. Const. amend. IV (explaining that in order to effectuate a reasonable search, one must obtain a warrant through probable cause supported by oath, granted by the courts).
[240] See supra Sections III.A-C.
[241] See supra Section IV.D.
[242] See supra Section V.A.
[243] See infra Section V.B.
[244] See infra Section V.B.
[245] See FlightHub 2 Platform, supra note 89 (providing a cloud management platform for drone surveillance procedures); discussion infra notes 246-47; see also DFR Public Dashboard, supra note 12 (providing logs and data of previously tracked drone surveillance paths).
[246] Carpenter v. United States, 585 U.S. 296, 313 (2018).
[247] Leaders of a Beautiful Struggle v. Balt. Police Dep’t, 2 F.4th 330, 346 (4th Cir. 2021) (en banc).
[248] Carpenter, 585 U.S. at 316; Leaders of a Beautiful Struggle, 2 F.4th at 346.
[249] MCPD Press Release, supra note 12; see also FlightHub 2 Platform, supra note 89 (providing a cloud management platform for drone surveillance procedures); DFR Public Dashboard, supra note 12 (providing logs and data of previously tracked drone surveillance paths).
[250] MCPD Press Release, supra note 12; see infra Section V.B.
[251] MCPD Press Release, supra note 12; see also FlightHub 2 Platform, supra note 89 (providing an account for storing surveillance); DFR Public Dashboard, supra note 12 (providing access to prior drone surveillance).
[252] See discussion infra notes 253-57.
[253] See Carpenter, 585 U.S. at 316; Leaders of a Beautiful Struggle, 2 F.4th at 346.
[254] See Carpenter, 585 U.S. at 316; Leaders of a Beautiful Struggle, 2 F.4th at 346.
[255] Lerussi, supra note 96.
[256] See Orin S. Kerr, Executing Warrants for Digital Evidence: The Case for Use Restrictions on Nonresponsive Data, 48 Tᴇx. Tᴇᴄʜ L. Rᴇᴠ. 1, 9 (2015) (arguing that subsequent use of lawfully seized digital data for a different investigative purpose is not inherently constitutional and should require renewed judicial authorization). See generally Emily Berman, When Database Queries Are Fourth Amendment Searches, 102 Minn. L. Rᴇᴠ. 577, 637-38 (2017) (claiming that developing technology requires new legal framework to combat new issue, including subsequent queries derived from over seized data).
[257] See supra Section V.B; Kerr, supra note 256, at 4. A commonly cited doctrinal foundation for secondary-use warrant requirements—particularly where law enforcement seeks to repurpose lawfully obtained data for a new investigative objective—derives from Fourth Amendment limits on scope, duration, and purpose. See Kerr, supra note 256, at 4 (stating that it “is time for courts to interpret the Fourth Amendment as imposing a use restriction on nonresponsive files seized during the execution of a warrant for digital evidence. Interpreting the Fourth Amendment as imposing a use restriction on nonresponsive data seized under a warrant can restore the needed limits on the warrant authority in light of the new facts of computer searches”).
[258] See infra Section V.C.
[259] U.S. Const. amend. IV; Md. Const. Decl. Of Rts., art 26.; Collins v. Virginia, 548 U.S. 586, 592-93 (2018) (citing California v. Ciraolo, 476 U.S. 207, 212–213 (1986)).
[260] Kyllo v. United States, 533 U.S. 27, 34 (2001) (quoting Silverman v. United States, 365 U.S. 505, 512 (1961)).
[261] Id. at 40.
[262] Id. at 36-38.
[263] See discussion and cases cited supra notes 258-62; see also infra Parts VI-VII (discussing suggested changes to Maryland’s statute to reflect drone surveillance, artificial intelligence, and supervised autonomy concerns).
[264] See discussion infra notes 265-69.
[265] See Julie E. Cohen, What Privacy Is For, 126 Harv. L. Rev. 1904, 1906 (2013) (emphasizing that mechanical or systemic controls can enforce constitutional norms).
[266] See Amy Sherry Fischer et al., Drones: A New Front in the Fight Between Government Interests and Privacy Concerns, 84 Def. Couns. J. 1, 4, 13 (2017); see, e.g., Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/15(6) (West 2023) (limiting law enforcement drone use to emergency or disaster-related purposes and implicitly requiring authorization for non-emergency surveillance).
[267] See discussion infra notes 268-69; infra Part VI.
[268] See supra Part IV.
[269] See supra Section V.A; infra Sections VI.A-B.
[270] See supra Part V; infra Part VI.
[271] See infra Part VI.
[272] See infra Part VI.
[273] See supra Sections V.A-C; infra Part VI.
[274] See generally Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/1 (West 2025) (demonstrating that Illinois has an existing statutory framework regulating police drone use).
[275] See infra Sections VI.A-C; Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/1, et seq. (West 2025).
[276] See Chicago the First City to Get Serious about Drone Injuries, Lerner & Rowe (Aug. 25, 2020), https://lernerandrowe.com/chicago-first-city-drone-laws/ (on file with the University of Baltimore School Law Forum); see also Anne Toomey McKenna & Clifford S. Fishman, Wiretapping and Eavesdropping, § 28:130. Illinois—Unmanned Aircraft Surveillance (2025) (discussing the Illinois Freedom from Drone Surveillance Act as comprehensive, because it regulates the retention and disclosure of collected data). However, several states enacted statutes regulating police drone use prior to Illinois, though those statutes did not establish comprehensive regulatory frameworks. See Va. S.B. 1331, 2013 Gen. Assemb., Reg. Sess. (approved Apr. 3, 2013) (2-year moratorium on drone use by law enforcement, except in emergencies); 2013 Idaho Laws Ch. 328 (S.B. 1134) (requiring police to obtain a warrant before conducting surveillance, except in emergencies); Fl. Stat. § 934.50 (2013) (prohibiting law enforcement from using drones to gather evidence except in specified circumstances); see also 2015 North Dakota Laws Ch. 239 (H.B. 1328) (imposing warrant and evidentiary restrictions on governmental drone surveillance, subject to limited statutory exceptions).
[277] See 725 Ill. Comp. Stat. Ann. 167/10, /15 (West 2025).
[278] See id. 167/10; id. 167/15(2)-(6).
[279] See id. 167/20.
[280] See discussion infra notes 281-84.
[281] See 725 Ill. Comp. Stat. Ann. 167/10, /15 (West 2025).
[282] See id. 167/17 (West 2025).
[283] Id. 167/25.
[284] See discussion and sources cited supra notes 281-83.
[285] See discussion infra note 286-87.
[286] U.S. Const. amend. IV; Md. Const. Decl. Of Rts., art. 26 (West 2025); Ill. Const. 1970, art. I, § 6 (West 2025).
[287] See discussion supra notes 285-86; infra Section VI.B.
[288] See infra Section VI.B.
[289] MCPD Press Release, supra note 12.
[290] See Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/5 (West 2025) (illustrating Illinois’s practice of expressly defining surveillance-related technology, including “drone,” as “any aerial vehicle that does not carry a human operator”).
[291] See supra Sections III.A-B.
[292] See supra Section V.A.
[293] See generally Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/20(b)(1)-(4) (West 2025) (authorizing retention of information under certain circumstances, which necessarily entails maintaining persistent, time-stamped operational logs).
[294] See discussion supra notes 233-57; see also Karen Kent et al., Guide to Integrating Forensic Techniques into Incident Response, NIST Spec. Publ. 800-86 at ES-1 (2006), https://nvlpubs.nist.gov/nistpubs/legacy/sp/nistspecialpublication800-86.pdf (on file with the University of Baltimore Law Forum) (explaining that documentation and logging are essential to maintaining chain of custody and evidentiary integrity in digital forensics).
[295] See Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/35 (West 2025) (reflecting legislative recognition that law-enforcement drone use requires standardized policies and public accountability mechanisms).
[296] See supra Section III.B (arguing that statutory governance must extend to the analytic models used in law enforcement flight systems); see, e.g., Md. Code Ann., Gen. Provisions, § 4-103 (West 2024) (“All persons are entitled to have access to information about the affairs of government and the official acts of public officials and employees.”).
[297] See Md. Code Ann., Gen. Provisions § 4-103 (West 2024); see also Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/25 (West 2025) (illustrating Illinois statutory formatting that Maryland could mirror under the Public Information Act).
[298] Compare discussion supra notes 290-97 (Maryland should adapt their statutory text to cover: (1) Explicit AI Coverage, (2) Human Confirmation and Audit Logging, (3) Oversight and Transparency, and (4) Integration with Existing Statutes), with Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/10 (West 2025) (prohibiting law enforcement use of drones to gather information, but not providing for operational safeguards, such as human confirmation or audit logging).
[299] See supra Section V.A.
[300] Marbury v. Madison, 5 U.S. 137, 163 (1803) (“The very essence of civil liberty certainly consists in the right of every individual to claim the protection of the laws, whenever he receives an injury.”).
[301] 725 Ill. Comp. Stat. Ann. 167/10, /30 (West 2025).
[302] See discussion infra notes 303-04; Mapp v. Ohio, 367 U.S. 643, 655 (1961).
[303] Mapp, 367 U.S. at 655.
[304] See id. at 657; discussion supra notes 184-203.
[305] See Alexander v. Sandoval, 532 U.S. 275, 286–87 (2001) (explaining that a private right of action exists only when Congress manifests an intent to create both a private right and a private remedy); discussion infra notes 306-07; infra Attachment A § 9; see also Cannon v. Univ. of Chi., 441 U.S. 677, 688–89 (1979) (discussing the role of private rights of action in ensuring statutory compliance); cf. Daniel J. Solove & Paul M. Schwartz, Privacy Law Fundamentals 121–23 (6th ed. 2022) (describing private rights of action as a critical accountability mechanism in privacy and surveillance statutes).
[306] Overview of the Privacy Act: 2020 Edition, Off. Of Priv. & C.L. https://www.justice.gov/opcl/overview-privacy-act-1974-2020-edition/remedies?utm_source=chatgpt.com (on file with the University of Baltimore Law Forum) (last visited Apr. 7, 2026).
[307] See Daphna Renan, The Fourth Amendment as Administrative Governance, 68 Stan. L. Rev. 1039, 1043–44 (2016) (arguing that vindicating Fourth Amendment values in an era of programmatic surveillance requires administrative governance mechanisms, including tools independent of criminal procedure).
[308] See discussion infra notes 309-14; infra Attachment A § 9.
[309] Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/45 (West 2025).
[310] See id.
[311] See supra Part IV.
[312] See infra Part VII.
[313] See supra Part VI (articulating the proposed statutory framework for autonomous drone governance).
[314] See infra Part VII.
[315] Understanding Policy Implementation, Implementation Guides (describing how implementation frameworks shape policy outcomes), https://implementation.effectiveservices.org/context/policy-implementation (on file with the University of Baltimore Law Forum) (last visited Apr. 1, 2026).
[316] See supra Part VI.
[317] See infra Part VII.
[318] See supra Part II.
[319] See supra Part III; infra Part VII.
[320] See infra Part VII.
[321] See discussion infra notes 322-23.
[322] See infra Part IV; Wilson v. Arkansas, 514 U.S. 927, 934 (1995) (holding that Fourth Amendment reasonableness includes consideration of the manner of execution of a search).
[323] See discussion infra notes 324-34.
[324] 14 C.F.R. § 107.19 (2025); 14 C.F.R. § 107.15 (2025); see, e.g. Advisory Circular 120-17B, Fed. Aviation Admin. (Dec. 19, 2018), https://www.faa.gov/documentLibrary/media/Advisory_Circular/AC_120-17B.pdf (on file with the University of Baltimore Law Forum).
[325] FAA, Advisory Circular 107-2B: Small Unmanned Aircraft Systems Operations § 7.3 (2024).
[326] See discussion and cases cited supra notes 324-25.
[327] See generally International AI Safety Report 2026 (warning that general-purpose AI systems can remain unreliable, produce inconsistent and incorrect outputs, and make it difficult to assess system behavior in real-world conditions, which poses safety and governance risks), International AI Safety Report (Feb 3, 2026), https://internationalaisafetyreport.org/publication/international-ai-safety-report-2026 (on file with the University of Baltimore Law Forum).
[328] See Md. Code Ann., Econ. Dev. § 14-301 (West 2024).
[329] Id.
[330] See id.; Jim Dempsey, Preemption of State Cybersecurity Laws: It’s Complicated, Lawfare (Aug. 24, 2022, at 08:01 ET), https://www.lawfaremedia.org/article/preemption-state-cybersecurity-laws-its-complicated (on file with the University of Baltimore Law Forum).
[331] See Md. Code Ann., Transp. § 5-1001(b)(2) (West 2024).
[332] See Econ. Dev. § 14-301 (West 2024) (conditioning state authorization of certain technological activities on compliance with statutory regulatory safeguards, thereby linking operational integrity to lawful deployment); Transp. § 5-1001(b)(2) (West 2024) (requiring compliance with statutory operational and safety standards as a prerequisite to lawful operation of regulated systems).
[333] See discussion supra notes 322-32.
[334] See discussion supra notes 321-33.
[335] See discussion infra notes 336-49.
[336] United States v. Leon, 468 U.S. 897, 922 (1984).
[337] See discussion infra notes 338-49.
[338] See discussion infra notes 339-41.
[339] Md. Code Ann., Pub. Safety § 3-506 (West 2025).
[340] Dawn Zoldi, Drone Training is Critical for Public Safety Now More Than Ever, Autonomy Glob. (Mar. 27, 2025), https://www.autonomyglobal.co/drone-training-is-critical-for-public-safety-now-more-than-ever/ (on file with University of Baltimore Law Forum).
[341] Mateusz Dolata & Gerhard Schwabe, Moving Beyond Privacy and Airspace: Guidelines for Just Drones in Policing, 40 Gov’t Info. Q. 1, 15 (2023); see supra Part V.
[342] See discussion infra notes 343-44.
[343] Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 587-88 (1993).
[344] See The Science Behind Breathalyzer Tests in Maryland DUI Cases, Alpert Schreyer Criminal Defense Attorneys (Mar. 19, 2024), https://andrewalpert.com/blog/the-science-behind-breathalyzer-tests-in-maryland-dui-cases/ (on file with University of Baltimore Law Forum) (highlighting the need to preserve records as a means to challenge reliability and accuracy on appeal).
[345] See discussion infra notes 346-48.
[346] See Thomas Black, Increasing Use of Drones by Police Will Need Rules, Herald Net (Sep. 24, 2025), https://www.heraldnet.com/opinion/comment-increasing-use-of-drones-by-police-will-need-rules/ (on file with the University of Baltimore School of Law); supra notes 213-16.
[347] See Robert Kane & Jordan Hyatt, Lessons from ‘Stop and Frisk’ Can Help Police Use Drones Without Compromising Civil Liberties, Route Fifty (Jan. 22, 2025), https://www.route-fifty.com/public-safety/2025/01/lessons-stop-and-frisk-can-help-police-use-drones-without-compromising-civil-liberties/402398/ (on file with the University of Baltimore Law Forum); supra Sections VI.B-C.
[348] See Prajwal Gowda, What Can Go Wrong When Compliance Stops at a One-Time Audit, Ampcus Cyber (Dec. 23, 2025), https://www.ampcuscyber.com/blogs/what-can-go-wrong-when-compliance-stops-at-a-one-time-audit/(on file with the University of Baltimore Law Forum).
[349] Dolata & Schwabe, supra note 341.
[350] Id.; see supra Part V.
[351] See Md. Code Ann., Transp. § 5-1001(a) (West 2024).
[352] See Freedom from Drone Surveillance Act, 725 Ill. Comp. Stat. Ann. 167/25(b) (West 2025).
[353] Marie-Rose Sheinerman, Police nationwide are embracing a new first responder: Drones, The Washington Post (Aug. 4, 2025), https://www.washingtonpost.com/nation/2025/08/04/drones-first-responders-police-departments/ (on file with the University of Baltimore Law Forum).
[354] Faine Greenwood, How to Regulate Police Use of Drones, Brookings Inst. (Sept. 24, 2020), https://www.brookings.edu/articles/how-to-regulate-police-use-of-drones/ (on file with the University of Baltimore Law Forum).
[355] See discussion infra notes 356-58.
[356] Md. Code Ann., Transp. § 5-104(b) (West 2024); Drone Security in Law Enforcement: Opportunities and Policy Challenges, Ares (Feb. 17, 2023), https://aressecuritycorp.com/2025/02/17/drone-security-in-law-enforcement-opportunities-and-policy-challenges/ (on file with the University of Baltimore Law Forum).
[357] Brittney D. Kohler et al., Drones Above Your City – How Cities Engage with Federal Officials on Drones, Nat. League of Cities (Jan. 3, 2025), https://www.nlc.org/article/2025/01/03/drones-above-your-city-how-cities-engage-with-federal-officials-on-drones/ (on file with the University of Baltimore Law Forum); see Ares, supra note 356; supra Section III.D.
[358] See discussion supra notes 349-57.
[359] See supra Part VI.
[360] See supra Sections VI.A-B.
[361] See infra Part VIII.
[362] See supra Part I.
[363] See discussion and sources cited supra notes 1-10.
[364] See discussion infra notes 365-66.
[365] Jamiles Lartey, It’s a Bird. It’s a Plane. It’s the Police Responding to 911, The Marshall Project (July 29, 2023, at 12:00 ET), https://www.themarshallproject.org/2023/07/29/police-drone-california-911#:~:text=In%20a%20report%20published%20on,of%20people%20who%20sign%20up (on file with the University of Baltimore Forum).
[366] See DFR Public Dashboard, supra note 12.
[367] See supra Part IV.
[368] See supra Parts IV-V.
[369] See supra Part VI.
[370] See discussion supra notes 46-50.
[371] See discussion supra notes 46-50.
[372] See supra Part VI.
[373] See supra Part VI.
[374] See supra Part VI.
[375] See supra Parts V-VI.
[376] See supra Section IV.B.
[377] See supra Part II.
[378] See supra Part II and Section III.D.
[379] See supra Section V.A.
[380] See supra Section IV.D.
[381] See supra Sections IV.A-B.
[382] See supra Sections IV.A-B.
[383] See supra Part VI.
[384] See supra Part VI.
[385] See supra Part VI.
[386] Drones with Artificial Intelligence: What Can We Expect in the Coming Years, HPDRONES (Oct. 24, 2025), https://hp-drones.com/en/drones-with-artificial-intelligence-what-can-we-expect-in-the-coming-years/ (on file with the University of Baltimore Law Forum).
[387] See supra Part VI.
[388] See supra Part VI.





Leave a Reply