Invisible Systems, Visible Consequences: Microscopic Drones, Transformative Military Power, and Catastrophic Risk

Invisible Systems, Visible Consequences: Microscopic Drones, Transformative Military Power, and Catastrophic Risk

By Christoph Bluth

1 Introduction: The Invisibility Paradox in Micro-Scale Robotics

Traditional aviation and security regulations are predicated on the physical detectability of aircraft as the basis for safety, control, and accountability. The emergence of micro-air vehicles (MAVs), however, challenges this foundation because these systems are designed to be extremely small, low-signature, and inherently covert (Davis et al., 1996, p. 197). Miniaturisation does not simply reduce scale; it fundamentally alters the structure of risk. As platforms move into the sub-gram domain, the primary concern shifts away from kinetic impact toward problems of detectability, attribution, and retrieval (Microscopic Drone Technologies: Mapping Potential Benefits, Societal Risks, and Governance Gaps Across Insect-Scale Robots, Smart Dust, and Micro/Nanorobots, 2024, p. 4). Operating with bird-like flight profiles and minimal acoustic signatures, such systems can evade conventional radar and monitoring architectures (Davis et al., 1996, pp. 198–199). This creates a structural visibility–governance gap in which the mechanisms that underpin existing regulatory regimes become ineffective. The implications are not merely technical but strategic: a class of systems is emerging that can operate below established thresholds of detection and control, thereby reshaping both civilian oversight and military capability. This paper examines how these characteristics require a shift away from aviation-centric models of governance toward frameworks grounded in traceability, accountability, and systemic risk.

1.1 Research Context, Scope, and Objectives

The development of microscopic drones represents a fundamental departure from conventional aviation, not only in terms of scale but in the underlying physical and operational principles that govern their use. This paper focuses on micro air vehicles (MAVs), typically defined by wingspans ranging from approximately 6 inches to 15 centimetres (Coffey & Montgomery, 2002, p. 1; Huber, 2002, pp. 2–5). At this scale, flight occurs within low Reynolds number regimes between 20,000 and 50,000, where aerodynamic behaviour is dominated less by inertia and more by viscous forces, fundamentally altering lift, drag, and control dynamics. These systems are not primarily driven by clearly articulated operational requirements, but by advances in micro-electromechanical systems (MEMS) and miniaturised electronics, which have enabled new classes of platforms to emerge (Huber, 2002, p. 2). While miniaturisation imposes severe engineering constraints, it simultaneously opens up operational possibilities that are inaccessible to larger systems, particularly in confined, high-risk, or denied environments. As a result, microscopic drones occupy a distinct niche characterised by proximity, persistence, and reduced detectability, rather than endurance or payload capacity. This study conceptualises microscopic drones as an umbrella category encompassing three distinct but interrelated technology families: insect-scale MAVs based on bio-inspired flight, smart dust systems composed of distributed millimetre-scale sensors, and medical or environmental nanorobots operating in fluid or biological environments (Microscopic Drone Technologies: Mapping Potential Benefits, Societal Risks, and Governance Gaps Across Insect-Scale Robots, Smart Dust, and Micro/Nanorobots, 2024, pp. 1–2). These systems differ fundamentally in their physical operating environments and functional design, yet share a common logic of extreme miniaturisation and high system integration (Huber, 2002, p. 14). Their strategic significance derives from the “power of proximity,” whereby very small payloads can achieve disproportionate effects by operating close to targets with minimal energy requirements (Coffey & Montgomery, 2002, p. 8). At the same time, it is essential to distinguish between experimental prototypes—such as sub-gram research platforms—and fully operational systems, as the gap between laboratory capability and field deployment remains substantial (Microscopic Drone Technologies…, 2024, p. 2). Treating these technologies as a unified category without such differentiation risks overstating both their current capabilities and their immediate policy implications.

The primary objective of this study is to analyse how advances in micro-scale robotics reconfigure the relationship between technical capability and societal risk, with particular emphasis on the shift from physically observable hazards to problems of detectability, attribution, and control (Microscopic Drone Technologies: Mapping Potential Benefits, Societal Risks, and Governance Gaps Across Insect-Scale Robots, Smart Dust, and Micro/Nanorobots, 2024, p. 4). To address this, the paper adopts a scoping review approach to synthesise fragmented evidence across engineering, medical, and legal domains, enabling a systematic comparison between demonstrated capabilities and projected applications (Microscopic Drone Technologies…, 2024, pp. 4–7). A central concern is the emergence of a persistent governance lag, in which rapid technological development outpaces the capacity of regulatory frameworks to respond effectively. By identifying key constraints in autonomy, range, and endurance, the analysis seeks to ground policy discussions in technical realities rather than speculative scenarios (Huber, 2002, p. 32; Coffey & Montgomery, 2002, pp. 6–8). In doing so, the study aims to bridge the gap between laboratory innovation and real-world deployment, providing a more robust basis for assessing both the transformative potential and the systemic risks of microscopic drone technologies.

1.2 Methodological Approach: Scoping Review and Multi-Level Evidence Mapping

To address the fragmented and interdisciplinary nature of research on microscopic drones, this study adopts a scoping review methodology. This approach is appropriate given that relevant evidence is dispersed across multiple domains, including robotics, nanomedicine, security studies, and regulatory analysis (Microscopic Drone Technologies: Mapping Potential Benefits, Societal Risks, and Governance Gaps Across Insect-Scale Robots, Smart Dust, and Micro/Nanorobots, 2024, p. 1). The review is guided by the PRISMA-ScR framework, which provides a structured basis for identifying, categorising, and synthesising diverse forms of evidence while maintaining methodological transparency (Microscopic Drone Technologies…, 2024, p. 1). Rather than treating technological development and governance as separate domains, the approach adopts a bottom-up logic in which policy-relevant insights are derived directly from engineering constraints and demonstrated capabilities (Drukarch et al., 2023, p. 3). This is particularly important in a field where speculative narratives frequently outpace verified technical progress, creating a risk of misaligned policy responses. The evidence base for this review is constructed through systematic desktop research drawing on academic databases such as IEEE Xplore and PubMed, complemented by policy reports and technical standards to capture both scientific and regulatory perspectives (Microscopic Drone Technologies…, 2024, pp. 4–7; Drukarch et al., 2023, p. 8). This mixed corpus allows the analysis to move beyond purely technical assessments by incorporating insights into how these systems are understood and governed in practice. To address gaps between formal regulation and real-world implementation, the methodology integrates iterative feedback mechanisms, including stakeholder-oriented discussions and workshop-based insights where available, in order to identify challenges that may remain underrepresented in formal literature (Drukarch et al., 2023, pp. 8–10). This combination of technical and qualitative sources ensures that the analysis reflects both the state of engineering development and the practical limitations of existing governance frameworks. During the evidence mapping stage, the literature is systematically categorised using variables such as physical scale, autonomy, and demonstrated operational capability in order to distinguish between experimentally verified systems and speculative or projected applications (Microscopic Drone Technologies…, 2024, pp. 2–4). A key analytical step is the separation of laboratory-based proof-of-concept platforms—such as sub-gram flight systems—from mature, field-deployable technologies, which remain significantly more limited in capability (Microscopic Drone Technologies…, 2024, p. 2). These technical characteristics are then linked to a multidimensional risk framework that evaluates factors such as likelihood, severity, detectability, and controllability, allowing for a structured comparison between engineering constraints and governance challenges (Microscopic Drone Technologies…, 2024, p. 3). This mapping process highlights the “pacing problem,” whereby the speed of technological development outstrips the adaptive capacity of legal and regulatory systems (Drukarch et al., 2023, pp. 3–5). By explicitly connecting hardware-level limitations with systemic risks, the methodology provides the analytical bridge necessary to assess how microscopic drones generate governance gaps that cannot be addressed through existing frameworks.

2. Taxonomy and Technology Families

This study categorizes microscopic robotic systems into three distinct technology families in order to address the persistent tendency in both policy and popular discourse to conflate these technologies despite their fundamentally different physical principles and operational implications. While all such systems share the defining characteristic of extreme miniaturisation at the sub-centimetre or millimetre scale, the aerodynamic constraints governing a flying platform differ radically from the sensing and communication architectures of distributed smart dust or the fluid dynamics shaping medical nanorobots. The classification framework applied distinguishes between aerial micro-platforms, surface-distributed sensor networks, and fluid-based micro-robotic systems, based on their primary mobility mechanisms and operational environments. This differentiation is analytically necessary because each category implies distinct capability profiles, developmental trajectories, and risk structures. The following sections examine these technology families in turn, focusing not only on engineering milestones but on the extent to which they translate into operational and, in particular, military capability.2.1 Insect-Scale Micro-Air Vehicles (MAVs): Engineering Milestones and Operational Realities.

The development of insect-scale micro-air vehicles represents a significant systems engineering challenge in which extreme miniaturisation imposes constraints that fundamentally reshape design choices and performance outcomes. At this scale, conventional components such as electromagnetic motors and rotary bearings become inefficient due to increased friction and unfavourable scaling effects, leading to a shift toward alternative actuation mechanisms (Jafferis et al., 2019, pp. 491–492). Piezoelectric actuators, in particular, are favoured because their power density improves as dimensions decrease, allowing them to generate thrust levels that can exceed those of biological muscle at comparable scales (Jafferis et al., 2019, p. 491). In parallel, non-mechanical propulsion concepts such as electrohydrodynamic (EHD) thrust have emerged, producing a silent “electric wind” through ion acceleration in a plasma field (Drew, 2018, pp. 6, 7–10). These approaches reflect a broader shift away from traditional aerodynamic solutions toward architectures that can function under the constraints imposed by low Reynolds number flight. While such innovations demonstrate that controlled flight at the sub-gram scale is technically feasible, they also highlight the extent to which capability is contingent on overcoming severe physical limitations, raising questions about the extent to which laboratory breakthroughs can be translated into robust operational systems.Recent milestones demonstrate that untethered flight is technically feasible even for sub-gram systems. The RoboBee X-Wing achieved a peak lift-to-weight ratio of 4.1:1 by utilizing a four-wing configuration, which increased lift efficiency by 29% over previous two-wing models (Jafferis et al., 2019, p. 491). Further miniaturization is evident in subcentimeter prototypes weighing only 21 mg that achieve navigable flight via external magnetic fields (Lin et al., 2025). These magnetic robots use a gyroscopic effect to stabilize flight attitude, achieving a survival rate of 76.5% during collision tests (Lin et al., 2025). Parallel to these lab demos, bionic designs like the mosquito-sized drones developed at the National University of Defense Technology are being tailored for battlefield reconnaissance (Mankel, 2025, pp. 1, 2). These drones feature stick-shaped bodies and leaf-like wings to ensure they remain nearly invisible to the naked eye during special missions (Mankel, 2025, p. 3). Such developments highlight a tactical shift toward using lightweight, thin components to facilitate covert operations (Mankel, 2025, pp. 1–3). Recent experimental milestones demonstrate that untethered flight at the sub-gram scale is technically achievable, but they also illustrate the narrow conditions under which such performance can be sustained. Platforms such as the RoboBee X-Wing achieve high lift-to-weight ratios through multi-wing configurations that enhance aerodynamic efficiency (Jafferis et al., 2019, p. 491). Further miniaturisation has produced sub-centimetre devices weighing as little as 21 mg, capable of controlled flight using externally applied magnetic fields to stabilise motion (Lin et al., 2025). At the same time, developments such as mosquito-sized reconnaissance drones indicate a growing interest in translating these technologies into operational contexts, particularly for surveillance and covert military applications (Mankel, 2025, pp. 1–3). These examples highlight the potential for microscopic drones to exploit their small size for proximity-based sensing and low observability. However, the reliance on external control systems, highly controlled environments, or specialised propulsion mechanisms underscores the gap between experimental capability and autonomous field deployment. The emerging military relevance of these systems therefore lies less in immediate operational effectiveness than in their prospective role as enablers of new forms of distributed, low-signature sensing and surveillance.

2.2 Smart Dust and Distributed Sensor Networks: Paradigms, Connectivity, and Ecological Integration

The smart dust paradigm envisages a distributed network of highly miniaturised, autonomous sensor nodes, often no larger than a grain of sand, capable of integrating sensing, computation, and communication within a few cubic millimetres (Mondal & Haick, 2025). These systems are enabled by advances in micro-electromechanical systems (MEMS) and ultra-low-power electronics, which allow for the embedding of multiple functions within extremely constrained physical volumes (Kahn et al., 1999, pp. 1–2). However, the feasibility of such systems is tightly bounded by severe energy constraints: a cubic-millimetre device typically stores on the order of one joule of energy, requiring continuous operation at power levels below approximately 10 microwatts to sustain functionality over meaningful time periods (Kahn et al., 1999, p. 1). Within these limits, individual nodes are relatively simple, but their collective behaviour enables the formation of self-organising networks capable of multi-hop communication and distributed data processing (Anderson et al., 2003, p. 1). The strategic significance of smart dust lies not in the capability of individual devices, but in the emergent properties of large-scale deployments, which have the potential to generate persistent, high-resolution sensing environments that are difficult to detect, attribute, or control. Communication within smart dust networks is shaped by the need to minimise energy consumption, leading to a preference for free-space optical transmission over conventional radio frequency systems, which require larger antennas and more complex modulation hardware (Kahn et al., 1999, pp. 2–4). Passive optical techniques, such as the use of corner-cube retroreflectors, enable low-power data transmission over moderate distances, but introduce significant constraints in terms of directionality and reliability (Kahn et al., 1999, p. 3). As a result, network architectures must operate under conditions of link asymmetry, where nodes may receive signals without being able to respond, rendering traditional bidirectional communication protocols ineffective (Kahn et al., 1999, p. 7). Within these constraints, swarm-level coordination emerges through adaptive routing and distributed decision-making, allowing large numbers of simple devices to generate complex spatial information, such as thermal or motion mapping, that exceeds the capability of individual nodes (Kahn et al., 1999, pp. 7–8). From a security perspective, this architecture creates both opportunity and vulnerability: while the distributed nature of the network enhances resilience and persistence, it also complicates control, attribution, and defence, particularly in scenarios where such systems are deployed covertly or at scale.

The large-scale deployment of smart dust systems raises significant challenges related to persistence, environmental impact, and long-term governance. Because these devices are typically dispersed in large numbers and are often non-retrievable, their accumulation introduces the risk of widespread electronic contamination in natural and built environments (Mondal & Haick, 2025; Kahn et al., 1999, pp. 3–5). To address this, research has increasingly focused on the development of biodegradable materials, such as cellulose, chitosan, and other organic substrates, which allow devices to degrade after completing their operational lifecycle (Mondal & Haick, 2025). At the same time, sustaining functionality prior to degradation requires alternative energy solutions, including wireless power transfer and ambient energy harvesting, which remain technically constrained at very small scales (Mondal & Haick, 2025). These trade-offs highlight a fundamental tension between persistence and disposability: systems designed for long-term sensing and monitoring inherently risk creating lasting environmental footprints, while those designed for rapid degradation may be limited in operational effectiveness. From a security perspective, the potential for large-scale, low-cost deployment of such networks introduces new forms of persistent, low-visibility surveillance and environmental monitoring, with implications that extend beyond technical feasibility into questions of control, accountability, and unintended systemic effects.

3. Technical Constraints: The Reality of Miniaturisation

Shrinking a robotic system is not simply a matter of reducing component size; it entails confronting physical constraints that increasingly dominate system behaviour at small scales. This section examines the technical reality of miniaturisation, where sub-gram platforms encounter diminishing returns as they move away from the assumptions that underpin conventional aviation and engineering design. A central challenge is the recursive interaction between size, weight, and power, which imposes a hard ceiling on achievable performance. At the same time, the very characteristics that make microscopic drones attractive—low mass and small form factor—also render them highly sensitive to environmental disturbances such as airflow, turbulence, and drag. These constraints define the operational envelope within which such systems can function. The following sections analyse these limitations in detail, focusing on energy and endurance, aerodynamic control at low Reynolds numbers, sensing and communication constraints, and the computational trade-offs associated with autonomous navigation. The key analytical point is that while miniaturisation enables new forms of capability, it simultaneously imposes structural limits that constrain their translation into robust, real-world systems.

3.1 SWaP Bottlenecks: Power, Weight, and Endurance

Miniaturisation in microscopic robotics is fundamentally constrained by a recursive design dynamic commonly described as the SWaP (size, weight, and power) spiral. Attempts to enhance functionality—by adding sensors, processing capability, or communication modules—generate a feedback loop in which increased power demand requires larger energy storage, which in turn increases system mass and reduces overall performance (Coppola et al., 2020, pp. 7, 9–10). These trade-offs are non-linear: higher current draw can degrade effective energy extraction from battery chemistry, further constraining endurance (Coppola et al., 2020, p. 7). As a result, micro-scale platforms are unable to integrate the full suite of capabilities required for autonomous operation without compromising flight duration or stability (Coppola et al., 2020, p. 3). This constraint is particularly acute in insect-scale systems, where actuation often requires high voltages and therefore relatively heavy onboard electronics, such as inductors, that consume a disproportionate share of the available payload capacity (James & Fuller, 2021, pp. 1–3). The implication is that increases in capability do not scale linearly, but instead encounter rapidly tightening physical limits. Energy storage remains the primary limiting factor for micro-scale flight, with battery mass often accounting for up to 30 per cent of the total system weight while still delivering only minutes of operational endurance (Cheah et al., 2019, p. 175; Coppola et al., 2020, p. 9). Conventional micro-battery designs are constrained by high proportions of inactive structural material, such as current collectors and binders, which reduce effective energy density at small scales (Wu et al., 2022, p. 2). Increasing power demand further exacerbates these limitations, as electrochemical constraints—particularly lithium-ion diffusion rates—restrict discharge performance under high loads (Wu et al., 2022, pp. 5–7). In parallel, the power electronics required to drive micro-actuators introduce additional inefficiencies, including heat generation and reactive losses, which further reduce usable energy (James & Fuller, 2021, pp. 1, 3). Taken together, these constraints define a hard ceiling on endurance that cannot be overcome through incremental improvements alone, but requires fundamentally different approaches to energy storage and management. Alternative energy and propulsion strategies offer potential pathways to mitigate battery constraints, but they introduce their own limitations. Electrostatic propulsion systems, such as the Coulombfly, demonstrate high lift-to-power efficiency and have enabled sustained flight under controlled conditions using solar energy (Shen et al., 2024). However, the effectiveness of solar harvesting declines rapidly with decreasing platform size, as the available surface area becomes insufficient to generate meaningful power (Coppola et al., 2020, p. 33). Wireless power transfer faces similar scaling challenges, with reduced receiver size leading to lower efficiency and increased thermal management requirements that exceed the payload capacity of micro-drones (Cheah et al., 2019, pp. 184–186). Moreover, many of these demonstrations remain subject to verification and are often achieved under highly controlled laboratory conditions, highlighting the gap between experimental feasibility and operational robustness (Shen et al., 2024). These approaches extend endurance in specific contexts, but they do not remove the fundamental trade-offs imposed by scale, reinforcing the conclusion that energy remains the central constraint shaping the capabilities of microscopic aerial systems. Alternative energy and propulsion strategies offer potential pathways to mitigate battery constraints, but they introduce their own limitations. Electrostatic propulsion systems, such as the Coulombfly, demonstrate high lift-to-power efficiency and have enabled sustained flight under controlled conditions using solar energy (Shen et al., 2024). However, the effectiveness of solar harvesting declines rapidly with decreasing platform size, as the available surface area becomes insufficient to generate meaningful power (Coppola et al., 2020, p. 33). Wireless power transfer faces similar scaling challenges, with reduced receiver size leading to lower efficiency and increased thermal management requirements that exceed the payload capacity of micro-drones (Cheah et al., 2019, pp. 184–186). Moreover, many of these demonstrations remain subject to verification and are often achieved under highly controlled laboratory conditions, highlighting the gap between experimental feasibility and operational robustness (Shen et al., 2024). These approaches extend endurance in specific contexts, but they do not remove the fundamental trade-offs imposed by scale, reinforcing the conclusion that energy remains the central constraint shaping the capabilities of microscopic aerial systems.

3.2 Aerodynamics and Control at Low Reynolds Numbers

Operating at chord Reynolds numbers between approximately 20,000 and 200,000, microscopic drones encounter aerodynamic conditions that differ fundamentally from those governing conventional aircraft. In this regime, viscous forces dominate over inertial effects, resulting in higher relative drag and reduced lift efficiency (Pelletier & Mueller, 2000, p. 825). Empirical studies indicate that cambered plate designs outperform flat surfaces by improving lift-to-drag ratios, while extremely thin wing profiles—often below 2 per cent thickness-to-chord ratio—are required to replicate the aerodynamic characteristics of biological flyers (Pelletier & Mueller, 2000, pp. 825–826). These design constraints reflect a shift away from traditional aerodynamic optimisation toward configurations that are specifically adapted to low-Reynolds-number flow. As a result, performance improvements depend less on scaling established designs and more on adopting fundamentally different aerodynamic principles. To compensate for reduced lift and efficiency at small scales, design strategies have focused on increasing effective wing area and exploiting resonance-based actuation. Multi-wing configurations, such as those employed in the RoboBee X-Wing, can improve lift generation by distributing aerodynamic load across a larger surface and operating at lower flapping frequencies (Jafferis et al., 2019, p. 492). Achieving stable flight under these conditions requires precise tuning of structural stiffness and transmission ratios to align with resonant frequencies, allowing systems to amplify useful motion while damping unwanted oscillations (Jafferis et al., 2019, pp. 492–494). Control strategies further exploit these properties by operating actuators near resonance, enhancing energy efficiency while maintaining predictable flight dynamics (Chen et al., 2019, p. 8; Jafferis et al., 2019, p. 493). These approaches demonstrate that controlled flight is achievable within the constraints of micro-scale aerodynamics, but only through tightly coupled design and control architectures that leave little margin for variation or disturbance.

The low mass of sub-gram platforms, while advantageous for reducing kinetic impact, introduces significant vulnerability to environmental disturbances. Even minor airflow variations, turbulence, or mechanical perturbations can destabilise flight, requiring high-frequency feedback control to maintain stability (Ho et al., 2024). In addition, aerodynamic loads can induce structural deformation or buckling in lightweight actuators, particularly in soft systems such as dielectric elastomer actuators (Chen et al., 2019, p. 8). To mitigate these effects, designers must balance rigidity and flexibility, often trading off efficiency against resilience. Soft actuators, for example, offer greater tolerance to collisions and deformation, but at the cost of lower energy efficiency and more complex control requirements. These trade-offs illustrate the broader challenge of maintaining stability in systems where the margin between controlled flight and failure is extremely narrow. Recent advances in control theory and machine learning have enabled improvements in manoeuvrability and disturbance rejection, but these gains are also constrained by hardware limitations. Techniques such as Robust Tube Model Predictive Control (RTMPC) allow micro-scale robots to execute complex manoeuvres and maintain trajectory tracking under moderate disturbances (Ho et al., 2024). However, the computational demands of such methods exceed the capabilities of most onboard processors, necessitating the use of simplified or learned approximations that can operate within strict memory and power limits. These approaches leverage the low inertia of micro-scale systems to achieve rapid response times, but they remain sensitive to modelling errors and environmental variability. As a result, high-performance control is achievable, but typically under controlled or semi-structured conditions rather than in fully open environments. Despite these advances, scaling penalties continue to impose significant limits on autonomy and operational robustness. Improvements in wing size or actuation efficiency can enhance lift and endurance, but often require additional power electronics and structural support, reintroducing the SWaP constraints discussed earlier (Chen et al., 2019, p. 13). Electrical losses in actuation systems, particularly in soft actuators, can dissipate a substantial proportion of input energy as heat, reducing overall system efficiency and limiting operational duration (Chen et al., 2019, pp. 23–25). Furthermore, small manufacturing asymmetries or minor deviations in component alignment can produce disproportionate effects on flight stability, highlighting the sensitivity of these systems to imperfections (Jafferis et al., 2019, pp. 494–496). Bridging the gap between laboratory demonstrations and reliable field deployment therefore requires not only advances in materials and control, but also improvements in manufacturing precision and system integration.

Taken together, these factors demonstrate that while micro-scale aerial platforms can achieve controlled and even agile flight under specific conditions, their performance remains tightly bounded by aerodynamic, structural, and computational constraints. The resulting capability profile is therefore highly asymmetric: these systems are well suited to short-duration, proximity-based sensing tasks in controlled environments, but remain limited in endurance, robustness, and autonomy in open settings. From a strategic perspective, their significance lies less in their immediate operational effectiveness than in their potential to enable distributed, low-signature sensing and surveillance architectures that exploit their small size and low detectability, while remaining dependent on external systems for coordination and control.

3.3 Sensing and Communication in Resource-Constrained Systems

Microscopic drones operate at what may be described as the “deep edge,” where computational, memory, and energy constraints are orders of magnitude below those of conventional edge devices. These systems typically function with only kilobytes of SRAM and low megahertz processing speeds, in contrast to the megabyte-level memory and gigahertz processors characteristic of standard platforms (Huckelberry et al., 2024, pp. 1–3). These constraints impose strict limits on onboard sensing, processing, and communication, requiring highly integrated system designs in which components perform multiple functions to minimise overhead (Davis et al., 1996, pp. 210–212). While it is technically feasible to incorporate relatively sophisticated sensors—even high-resolution imaging systems—such integration is contingent on extremely tight power budgets, often below tens of milliwatts (Davis et al., 1996, p. 210). As a result, sensing capability is not absent at the micro-scale, but tightly bounded and highly selective, favouring short-duration, task-specific data collection over continuous, high-bandwidth operation. Communication represents a central bottleneck, as extending range or data rate directly increases energy demand and system weight. To mitigate this, micro-scale platforms frequently rely on free-space optical communication, which avoids the need for large antennas and complex radio-frequency circuitry (Kahn et al., 1999, pp. 2–3). Passive optical techniques, such as corner-cube retroreflectors, enable low-power data transmission over distances of up to hundreds of metres, but introduce constraints in directionality, alignment, and reliability (Kahn et al., 1999, p. 3). Alternative approaches, including high-frequency radio links, offer greater range but at significantly higher energy cost, reinforcing the fundamental trade-off between communication capability and endurance (Davis et al., 1996, pp. 204–207). This trade-off is structurally similar to the SWaP spiral: increasing communication performance requires more power, which increases mass and reduces flight time, thereby constraining overall system effectiveness. At the network level, these constraints produce architectures that differ fundamentally from conventional communication systems. Micro-scale networks often operate under conditions of link asymmetry, where nodes may receive signals but lack the power or orientation to transmit responses, rendering standard bidirectional routing protocols ineffective (Kahn et al., 1999, pp. 6–7). Effective communication therefore relies on specialised protocols optimised for intermittent connectivity, directional transmission, and short, high-efficiency data bursts (Kahn et al., 1999, p. 5). Within these constraints, distributed networks can exhibit emergent behaviour, enabling large numbers of simple nodes to collectively generate complex spatial or environmental information (Kahn et al., 1999, pp. 7–8). This form of swarm-based sensing represents a qualitatively different capability: rather than relying on individual platform performance, it derives effectiveness from scale, redundancy, and coordination.

However, the integration of sensing, communication, and computation within such constrained environments introduces significant vulnerabilities, particularly in relation to security and system integrity. Standard security protocols, such as Transport Layer Security (TLS), are typically infeasible due to their memory and processing requirements, forcing reliance on lightweight alternatives that provide reduced protection (Huckelberry et al., 2024, pp. 10–12). In practice, this creates a direct trade-off between functionality and security, as computational resources devoted to machine learning or sensing cannot be used for encryption or authentication. The result is a class of systems that are inherently vulnerable to interception, manipulation, or spoofing, particularly when deployed in large numbers or in adversarial environments. At the swarm level, these vulnerabilities are amplified, as the compromise of individual nodes can propagate through the network, potentially degrading or corrupting collective behaviour. These constraints define a capability profile that is both limited and potentially transformative. On the one hand, individual micro-drones are severely restricted in terms of processing power, communication range, and security robustness. On the other hand, their ability to operate as distributed, low-signature sensing networks creates new possibilities for persistent, large-scale data collection that are difficult to detect or counter. From a strategic perspective, the significance of these systems lies not in their standalone performance, but in their potential to enable new forms of surveillance, reconnaissance, and environmental monitoring that exploit their small size, low cost, and scalability. At the same time, their inherent vulnerabilities and dependence on constrained resources highlight the fragility of these capabilities, reinforcing the broader conclusion that technological potential and operational reality remain tightly coupled at the micro-scale.

3.4 Navigation and Autonomy: Computational Trade-offs and Field Realities

Achieving reliable navigation and autonomy at the micro-scale remains one of the most significant technical challenges, as conventional approaches to localisation and control do not scale down effectively. Satellite-based navigation systems such as GPS are generally impractical for sub-gram platforms due to their power requirements and antenna size, which exceed the available energy and space budgets (Davis et al., 1996, pp. 200–207). Early experimental systems have therefore relied heavily on external infrastructure, including motion capture systems and ground-based tracking, to provide positional data and control inputs (Davis et al., 1996, p. 200). While such approaches demonstrate the feasibility of controlled flight, they do not translate into independent operation in open environments. At the same time, onboard alternatives such as inertial measurement units suffer from high drift rates at small scales, limiting their usefulness for sustained navigation (Davis et al., 1996, p. 207). As a result, absolute positioning remains a major unresolved constraint, with existing solutions providing only partial or context-dependent capability.

More advanced techniques, such as visual odometry and simultaneous localisation and mapping (SLAM), offer richer environmental awareness but are computationally intensive and therefore difficult to implement on highly constrained hardware (Coppola et al., 2020, p. 13). To address this, research has focused on simplified perception and control architectures that rely on limited sensor inputs, such as optical flow, combined with lightweight algorithms for local navigation (Zhou et al., 2022). These approaches enable micro-drones to operate in confined or structured environments, including narrow corridors or cluttered spaces, where relative rather than absolute positioning is sufficient. However, this comes at the cost of reduced situational awareness and increased dependence on environmental conditions, such as lighting and surface texture. The transition from externally supported systems to fully autonomous platforms therefore remains incomplete, with current solutions representing partial adaptations rather than fully generalisable capabilities.

The inherent instability of micro-scale flight further complicates the development of autonomous systems. At low mass and high actuation frequencies, small perturbations can lead to rapid deviations in trajectory, requiring continuous, high-rate feedback control to maintain stability (Davis et al., 1996, p. 205; Ho et al., 2024). Unsteady aerodynamic effects, including vortex formation and transient flow dynamics, introduce additional uncertainty that is difficult to model accurately in real time (Ho et al., 2024). Advanced control frameworks, such as Robust Tube Model Predictive Control, have demonstrated the ability to manage these dynamics and enable complex manoeuvres, but their computational requirements exceed the capacity of most onboard processors (Ho et al., 2024). As a result, practical implementations often rely on learned approximations or simplified control policies, which can operate within hardware limits but may be less robust under changing conditions. This highlights a recurring theme: improvements in capability are closely tied to—and limited by—computational constraints.

Collision avoidance and multi-agent coordination introduce further layers of complexity. Classical approaches, such as the Velocity Obstacle method, provide a theoretical basis for safe navigation in dynamic environments, but require accurate state estimation and continuous computation (Coppola et al., 2020, p. 24). At the micro-scale, these requirements are difficult to meet without external support or significant simplification. Some platforms compensate by incorporating passive resilience, such as protective structures or compliant materials, but these solutions increase weight and aerodynamic drag, thereby reducing overall efficiency (Coppola et al., 2020, pp. 15–17). More recent approaches emphasise decentralised planning and coordination, allowing swarms of micro-drones to navigate complex environments collectively without centralised control (Zhou et al., 2022). While this reduces the computational burden on individual nodes, it introduces new challenges related to coordination, communication, and fault tolerance.

The scalability of such systems depends critically on the balance between decentralisation and coordination. Centralised control architectures become increasingly inefficient as the number of agents grows, while fully decentralised systems require robust mechanisms for local decision-making and fault detection (Coppola et al., 2020, pp. 5–6, 34). Experimental demonstrations of swarm behaviour in uncontrolled environments suggest that such approaches are feasible, but remain sensitive to communication constraints and environmental variability (Zhou et al., 2022). The transition from laboratory conditions to real-world deployment therefore involves not only technical challenges, but also questions of reliability, redundancy, and system-level integration. These constraints indicate that fully autonomous, micro-scale aerial systems remain at an early stage of development. While significant progress has been made in individual components—such as sensing, control, and coordination—the integration of these capabilities into robust, independent systems capable of operating in complex environments remains limited. From a strategic perspective, this suggests that the near-term role of microscopic drones will depend less on full autonomy than on hybrid architectures, in which micro-scale platforms operate as components within larger, networked systems that provide navigation, control, and data processing support. Their significance therefore lies not in replacing existing platforms, but in augmenting them, enabling new forms of distributed sensing and low-signature operation that exploit their unique physical characteristics while remaining dependent on external infrastructure for effective deployment.

4. Societal and Security Risks

The transition from laboratory-scale innovation to real-world deployment exposes the central governance problem associated with microscopic drones: as systems decrease in size, they cross a threshold beyond which traditional mechanisms of detection, attribution, and control become ineffective. This paper argues that the defining feature of micro-scale robotic systems is not simply their technical novelty, but their capacity to operate below the visibility thresholds on which existing regulatory and security frameworks depend. This creates a structural visibility–governance gap in which the ability to deploy, observe, and intervene becomes asymmetrically distributed. Devices that are effectively invisible to conventional radar, human perception, or regulatory classification can bypass the categories that underpin aviation law, privacy regulation, and security oversight. The resulting shift is from visible, discrete risks—such as collision or physical damage—to diffuse and systemic vulnerabilities associated with surveillance, data extraction, and loss of control. To capture these dynamics, this section develops a four-part framework, examining the implications of invisibility for surveillance and attribution, the cybersecurity vulnerabilities of resource-constrained systems, the proliferation risks associated with dual-use technologies, and the environmental consequences of non-retrievable micro-scale devices.

4.1 The Surveillance Threat: Invisibility, Data Correlation, and the Attribution Void

Microscopic drones fundamentally alter the nature of surveillance by enabling forms of observation that are both persistent and effectively undetectable. Unlike conventional platforms, whose presence is typically visible or audible, micro-scale systems can operate without perceptible signatures, thereby removing the possibility of situational awareness for those being observed (Cavoukian, 2012, p. 13). This transforms surveillance from a visible, negotiated activity into an invisible and unilateral process, undermining established norms of consent and accountability. The combination of mobility, proximity, and miniaturisation allows these systems to collect data from vantage points that were previously inaccessible or prohibitively costly, enabling continuous monitoring over extended periods (Cavoukian, 2012, pp. 2–4). As a result, the risk is not simply an incremental increase in surveillance capability, but a qualitative shift toward environments in which observation can occur without detection, notification, or recourse. The implications extend beyond visual surveillance to the broader ecosystem of sensor data generated by micro-scale systems. Even in the absence of cameras, onboard sensors such as accelerometers, gyroscopes, and ambient light detectors can produce highly sensitive information, enabling the reconstruction of movement patterns, behavioural routines, and even security-relevant data such as login credentials (Kröger, 2019, pp. 148–154). When combined with external data sources, these streams can be aggregated into detailed and continuous profiles of individuals or environments. The key issue is that such data collection often occurs outside existing regulatory frameworks, which focus primarily on explicitly identifiable personal data. At the micro-scale, the distinction between innocuous and sensitive data becomes increasingly blurred, as seemingly low-value signals can be combined to produce high-value intelligence. This expands the scope of surveillance risk beyond traditional categories and complicates efforts to regulate data collection and use.

A central consequence of this shift is the emergence of an attribution gap, in which the operators of surveillance systems cannot be readily identified or held accountable. As Wang et al. (2016, pp. 180–184) note, drones function as extensions of remote operators, but at micro-scale the physical link between observer and observed is effectively severed. Even where formal registration systems exist, they provide no practical means for real-time identification, leaving individuals unable to determine who is collecting data or for what purpose. At the same time, technical features such as device fingerprinting enable operators to track targets persistently without revealing their own identity (Kröger, 2019, pp. 153–155). This asymmetry reverses traditional accountability structures: the subject becomes transparent, while the observer remains opaque. The result is a structural imbalance in which the capacity to observe is decoupled from the obligation to justify or regulate that observation. The combined effect of invisibility, data aggregation, and attribution failure is a breakdown of the conditions under which privacy can be meaningfully exercised. Conventional mechanisms—such as awareness, consent, or the ability to challenge surveillance—depend on the visibility of both the observing system and its operator. In the absence of these conditions, individuals cannot negotiate or enforce their rights, and surveillance becomes effectively non-consensual by default (Cavoukian, 2012, pp. 11–13; Wang et al., 2016, p. 178). This represents a fundamental governance challenge: existing legal and social frameworks assume that observation is detectable and attributable, assumptions that no longer hold at the micro-scale. As a result, microscopic drones do not simply extend existing surveillance practices, but undermine the foundational principles on which their regulation depends.

4.2 Cybersecurity in Micro-Scale Robotics: Resource Gaps, Swarm Fragility, and Lightweight Defence

Microscopic drones operate within a technical regime defined by extreme resource constraints, often described as the “deep edge,” where limited memory, processing power, and energy availability fundamentally restrict the implementation of conventional cybersecurity measures. While standard edge devices typically rely on megabytes of memory and relatively powerful processors, micro-scale systems are often limited to only a few kilobytes of SRAM and low-frequency microcontrollers, rendering established security architectures impractical (Huckelberry et al., 2024, pp. 1, 11–13). This constraint is not merely a technical inconvenience but a structural vulnerability, as it forces designers to prioritise core functionality—such as sensing, communication, or control—at the expense of robust security. As a result, many microscopic systems are deployed with minimal or incomplete protection, despite operating in environments where they may be physically accessible or exposed to adversarial interference.The incompatibility of standard security protocols with micro-scale hardware is particularly evident in the case of widely used encryption and authentication frameworks. Protocols such as Transport Layer Security (TLS) and Datagram TLS (DTLS), which form the backbone of secure communication in conventional networks, require memory and computational resources far beyond the capacity of typical microcontrollers used in microscopic drones (Huckelberry et al., 2024, pp. 11–13; Liu & Chakkaravarthy, 2026, p. 20). Even when lighter variants are employed, they often introduce latency or processing overheads that are incompatible with real-time operation. Consequently, developers must rely on simplified or lightweight cryptographic primitives, which provide only partial protection and may be vulnerable to interception, spoofing, or replay attacks. This creates a systemic trade-off in which the very constraints that enable miniaturisation simultaneously weaken the integrity and resilience of the system.

At the level of distributed systems, these vulnerabilities are amplified by the reliance on swarm-based architectures. Microscopic drones frequently operate as part of coordinated networks in which individual nodes share information and collectively generate system-level behaviour. While this enhances functionality and resilience to individual node failure, it also creates new attack surfaces. A compromised or malicious node can inject false data into the network, potentially influencing the behaviour of the entire swarm and degrading its effectiveness (Liu & Chakkaravarthy, 2026, p. 16). Given the limited capacity for real-time verification or anomaly detection at the node level, such attacks may be difficult to identify or mitigate once deployed. In addition, the compression and optimisation of machine learning models for micro-scale deployment can introduce hidden vulnerabilities, including backdoors that may only become active under specific conditions (Huckelberry et al., 2024, pp. 14–18). These characteristics highlight the fragility of trust in distributed micro-scale systems.

Efforts to mitigate these risks have focused on the development of lightweight and stratified security architectures that distribute protective functions across different layers of the system. For example, isolating highly constrained sensor nodes from more capable edge or gateway devices can reduce the overall attack surface while allowing more robust security measures to be implemented at higher levels of the network (Liu & Chakkaravarthy, 2026, p. 21). Decentralised trust mechanisms, such as Merkle-directed acyclic graphs, have also been proposed to enable scalable authentication and data integrity checks without the overhead of traditional consensus protocols (Liu & Chakkaravarthy, 2026, pp. 18–20). While such approaches offer promising avenues for improving resilience, they do not eliminate the underlying constraints, and their effectiveness depends on careful system design and integration. From a strategic perspective, the cybersecurity profile of microscopic drones is therefore characterised by a tension between scale and control. Their small size, low cost, and potential for large-scale deployment make them attractive for applications requiring distributed sensing and coordination, including military and intelligence operations. At the same time, their limited computational capacity and reliance on lightweight security mechanisms create inherent vulnerabilities that may be exploited by adversaries. This duality reinforces the broader argument of this paper: the capabilities enabled by miniaturisation are real, but they are inseparable from structural limitations that constrain reliability, security, and control. Microscopic drones operate within a technical regime known as the deep edge, where extremely limited memory and clock speeds create a fundamental mismatch with standard security protocols. While traditional edge devices typically utilize megabytes of RAM, micro-scale units are often restricted to a few kilobytes of SRAM, rendering conventional protection layers like heavy encryption or complex operating systems impractical (Huckelberry et al., 2024, pp. 1, 11–13). This resource gap is critical because the broad deployment of these platforms in diverse, physically accessible locations makes them inherently vulnerable to side-channel and fault injection attacks that target model weights (Huckelberry et al., 2024, p. 1). Furthermore, a significant research disparity exists in current literature, where engineering efforts for model optimization vastly outweigh dedicated security vetting (Huckelberry et al., 2024, pp. 2–4). Consequently, many microscopic systems are deployed with unvetted architectures that may not withstand sophisticated hardware-level exploitation. Standard internet security protocols such as Transport Layer Security (TLS) and Datagram TLS (DTLS) are often physically incompatible with the hardware limits of microscopic drones. The memory footprint of TLS can reach up to 300 KB, which far exceeds the 2 KB to 8 KB SRAM limits typical of many 8-bit microcontrollers (Huckelberry et al., 2024, pp. 11–13; Liu & Chakkaravarthy, 2026, p. 20). In empirical tests, attempts to execute full TLS on constrained hardware resulted in system failure, while DTLS exhibited prohibitive handshake latency (Liu & Chakkaravarthy, 2026, p. 20). Furthermore, hardware-level defenses such as masking are often unviable, as they can consume seven times more compute cycles than available (Huckelberry et al., 2024, p. 9). While lightweight primitives like AES and ECDH offer feasible alternatives, they require careful integration to ensure they do not compete with real-time machine learning inference for a drone’s limited energy and memory reserves (Huckelberry et al., 2024, pp. 2, 11–13).

4.3 Dual-Use and Proliferation: From Medicine to Warfare

The development of microscopic drones is inherently characterised by a dual-use dynamic in which advances in civilian research simultaneously expand the potential for military and malicious applications. Unlike technologies such as nuclear weapons, which require substantial industrial infrastructure and state-level coordination, micro-scale robotic systems can be assembled from commercially available components and open-source software, significantly lowering barriers to entry (Righetti & Boulanin, 2026, pp. 1–4). This diffusion pathway is accelerated by the culture of openness that underpins much of contemporary robotics research, where design architectures, control algorithms, and fabrication methods are widely shared. As a result, the boundary between legitimate scientific development and potential weaponisation becomes increasingly difficult to define, and the traditional distinction between civilian and military technology is eroded. Microscopic drones are particularly susceptible to weaponisation because effective applications do not necessarily require high levels of sophistication. In contrast to advanced military platforms, which depend on complex integration and high-end components, micro-scale systems can achieve meaningful operational effects through scale, proximity, and persistence rather than individual performance. This has led to concerns that autonomous or semi-autonomous systems could be deployed in large numbers to conduct surveillance, disrupt infrastructure, or deliver payloads with limited cost and technical expertise (Zając, 2025, pp. 3–4; King, 2023, p. 9). Such systems may favour the attacker by enabling low-cost, distributed operations that are difficult to detect or defend against, potentially altering the tactical balance in specific contexts. At the same time, their limited payload capacity and fragility constrain their ability to replace conventional military capabilities, indicating that their role is more likely to be complementary than transformative in the short term.

The accessibility of these technologies also has implications for non-state actors and smaller states, which may seek to exploit micro-scale systems to offset conventional military disadvantages. The use of commercially available drone technology by armed groups in recent conflicts demonstrates how rapidly such capabilities can diffuse and be adapted for hostile purposes (Righetti & Boulanin, 2026, p. 2). Microscopic systems extend this trend by enabling more covert and persistent forms of operation, including surveillance, harassment, or targeted disruption. However, their effectiveness remains context-dependent, as environmental conditions, limited endurance, and communication constraints restrict their operational reliability. In addition, these systems do not provide the capacity to control territory or achieve decisive outcomes in conventional warfare, reinforcing their role as enablers rather than substitutes for existing forms of military power. At the state level, proliferation dynamics are shaped by both technological diffusion and strategic competition. States may seek to develop or acquire micro-scale capabilities as part of broader efforts to enhance surveillance, intelligence gathering, and precision operations, particularly in contested or denied environments. At the same time, the transfer of such technologies to proxy actors or partner forces can serve as a means of extending influence while limiting direct involvement (O’Neill et al., 2024, pp. 17–23). This creates a complex landscape in which capabilities are dispersed across multiple actors, often without clear lines of accountability or control. Traditional export control regimes, which focus on identifiable hardware and well-defined categories, are poorly suited to managing these dynamics, particularly where critical components consist of software, data, or human expertise rather than physical systems (Székely, 2024, pp. 333–337).

Efforts to regulate proliferation therefore face significant challenges. The speed of technological development, combined with the modular and distributed nature of micro-scale systems, makes it difficult to identify and control relevant technologies before they are widely disseminated. Proposals such as tiered openness, gated access to sensitive data, and restrictions on high-risk applications represent attempts to balance innovation with risk mitigation (Righetti & Boulanin, 2026, pp. 5–6). However, these measures rely heavily on voluntary compliance and are unlikely to prevent determined actors from acquiring or adapting relevant capabilities. As a result, the governance of micro-drone proliferation must shift from controlling specific technologies to managing the conditions under which they are used, including the development of norms, accountability mechanisms, and defensive countermeasures. From a strategic perspective, the proliferation of microscopic drones does not fundamentally alter the balance of power at the level of great power competition, but it does introduce new forms of risk at lower levels of conflict. Their capacity for covert deployment, distributed operation, and persistent sensing creates opportunities for harassment, disruption, and targeted violence that are difficult to attribute and respond to effectively. This reinforces the broader pattern identified in this study: while micro-scale systems do not replace existing military capabilities, they expand the range of possible actions below the threshold of conventional warfare, complicating deterrence and increasing the potential for instability in already contested environments.

4.4 Ecological Persistence and the Toxicological Risks of Non-Retrievable Micro-Systems

The large-scale deployment of microscopic sensing systems introduces a distinct category of environmental risk arising from their persistence, dispersal, and limited retrievability. Smart dust architectures in particular rely on the distribution of vast numbers of micro-scale devices across natural or built environments, often in ways that make systematic recovery technically or economically unfeasible (Kahn et al., 1999, pp. 3–6). As devices shrink to sub-millimetre dimensions, they may remain suspended in air currents or embedded in soil and water systems, effectively creating a new class of distributed, and potentially permanent, electronic residue (Kahn et al., 1999, p. 1). This shifts the environmental impact of robotics from concentrated, recoverable systems to diffuse, cumulative contamination, in which the scale of deployment rather than the characteristics of individual devices becomes the primary driver of risk. The invisibility of these systems exacerbates the problem, as their small size makes tracking, monitoring, and retrieval extremely difficult once deployed (Mondal & Haick, 2025). In applications such as environmental monitoring or agriculture, where large numbers of devices are intentionally dispersed, the loss of a significant proportion of nodes is not an anomaly but an expected outcome. This creates a structural condition in which electronic waste accumulates as a by-product of normal operation. Where devices contain non-degradable materials, this raises concerns about long-term ecological effects, including the release of toxic substances into soil and water systems and potential harm to biological organisms (Mondal & Haick, 2025). The risk is therefore not limited to isolated incidents, but arises from the cumulative effects of repeated deployment at scale.

Efforts to address these challenges have focused on the development of biodegradable and transient electronics, using materials such as cellulose, chitosan, and other organic substrates that can decompose without leaving harmful residues (Mondal & Haick, 2025). These approaches aim to align the operational lifecycle of the device with its environmental impact, ensuring that functionality is maintained for the duration of the mission before degradation occurs. However, this introduces a fundamental design tension: materials that degrade too slowly risk accumulation and long-term contamination, while those that degrade too quickly may compromise device performance and reliability. In addition, environmental conditions—such as humidity, temperature, and biological activity—can significantly affect degradation rates, making it difficult to design systems with predictable lifecycles across different contexts. From a strategic perspective, the environmental dimension of microscopic drone deployment extends beyond ecological considerations to issues of control, persistence, and unintended consequences. Systems designed for large-scale sensing or surveillance may remain active, or at least physically present, long after their intended use, creating residual infrastructures that are difficult to regulate or remove. In military or security contexts, this raises the possibility of persistent sensing environments that continue to collect data or influence behaviour even after active operations have ceased. More broadly, the accumulation of non-retrievable micro-devices challenges existing regulatory frameworks, which are typically designed for discrete, identifiable technologies rather than distributed, transient systems. As with other aspects of microscopic drone technology, the key issue is not simply the existence of environmental risk, but the difficulty of managing that risk within existing systems of governance that rely on visibility, traceability, and control.

5 Governance Gaps and Regulatory Frameworks

The rapid miniaturisation of robotic systems exposes a fundamental limitation in existing governance frameworks: regulation remains anchored in assumptions about visibility, scale, and physical presence that no longer hold at the micro-scale. As demonstrated in previous sections, microscopic drones operate below the thresholds on which aviation law, privacy regulation, and security oversight are based, creating a structural mismatch between technological capability and legal control. This is not simply a question of regulatory delay, but a deeper problem of conceptual misalignment, in which the categories used to define and govern technology fail to capture its operational characteristics. Technological change alters social practices and risk perceptions faster than legal systems can adapt, producing a persistent governance gap. The central issue is therefore whether existing frameworks can be extended to accommodate these systems, or whether fundamentally new approaches are required. This section examines these challenges across four dimensions: the limitations of current legal regimes, the potential of Privacy by Design as a technical response, the role of Responsible Research and Innovation in managing dual-use risks, and the need for value-centered frameworks to bridge the gap between technological development and real-world application.

5.1 Limitations of Existing Aviation and Privacy Laws

The limitations of existing aviation and privacy laws reflect a broader problem of regulatory inertia in the face of rapid technological change. Legal frameworks are typically designed to provide stable and generalisable rules, but this stability becomes a liability when the underlying technological assumptions shift. In the case of microscopic drones, current regulations rely on categories—such as platform size, weight, and visibility—that no longer correspond to the operational reality of the systems they are intended to govern (Drukarch et al., 2023, pp. 2–5). This creates a structural disconnect in which devices fall outside existing classifications, leaving both regulators and developers operating in conditions of uncertainty. As a result, governance becomes reactive rather than anticipatory, addressing risks only after they have materialised. This mismatch is particularly evident in aviation regulation, which remains largely based on physical characteristics such as maximum take-off mass. Under existing frameworks, including those developed by the European Union Aviation Safety Agency, unmanned systems are classified into categories that determine operational restrictions and safety requirements (European Union Aviation Safety Agency, 2024). Microscopic drones frequently fall below the minimum thresholds for these categories, effectively placing them outside the scope of meaningful regulation. While this may reduce compliance burdens, it also removes safeguards, allowing systems with significant surveillance or security implications to operate without clear oversight. The underlying problem is that weight-based classification assumes a direct relationship between size and risk, an assumption that is reversed at the micro-scale, where smaller systems may pose greater challenges in terms of detectability and accountability. Privacy regulation faces a parallel difficulty. Frameworks such as the General Data Protection Regulation are built around principles of transparency, consent, and accountability, all of which depend on the ability of individuals to perceive and understand when data collection is taking place. The invisibility of micro-scale systems undermines these assumptions by enabling data collection without awareness, effectively bypassing the mechanisms through which consent is negotiated and enforced. Even where legal obligations exist, they cannot be meaningfully exercised if individuals are unable to detect the presence of the system or identify its operator. This creates a situation in which compliance becomes formally possible but practically irrelevant, highlighting the gap between legal principles and technological reality.

Similar limitations are evident in export control regimes, which struggle to address the dual-use nature of microscopic robotics. Traditional approaches focus on controlling the transfer of identifiable hardware or specific technologies, but micro-scale systems are often assembled from widely available components and depend heavily on intangible assets such as software, data, and expertise (Székely, 2024, pp. 333–337). As a result, attempts to regulate proliferation through list-based controls are easily circumvented, and enforcement becomes increasingly difficult. The same problem arises in relation to liability, where existing models assume clear chains of responsibility involving identifiable actors. In distributed or autonomous systems, particularly those operating as swarms, assigning responsibility for failure or harm becomes highly complex, raising fundamental questions about how accountability should be defined and enforced. A further challenge arises in the certification of autonomous micro-systems, where the absence of established standards creates a “chicken-and-egg” problem. Systems cannot be deployed without assurance of safety and reliability, but such assurance often requires data that can only be obtained through large-scale deployment (Computing Community Consortium, 2020, pp. 8–9). This circularity limits the ability of regulators to set meaningful requirements and of developers to demonstrate compliance. As a result, the governance of microscopic drones remains fragmented and incomplete, characterised by gaps in coverage rather than coherent oversight. The central implication is that existing frameworks, while still relevant in principle, are insufficient in practice, and that addressing the challenges posed by micro-scale systems requires a shift from adapting existing rules to rethinking the foundations on which they are based.

5.2 Privacy by Design (PbD) for Micro-Scale Systems: Technical Safeguards and Operational Standards

In response to the surveillance and accountability gaps identified in previous sections, Privacy by Design (PbD) offers a framework that seeks to embed privacy considerations directly into the technical architecture of micro-scale systems rather than relying on ex post regulatory enforcement (Cavoukian, 2012, p. 17; European Union Aviation Safety Agency, 2024). At its core, PbD shifts the focus from compliance to design, requiring that privacy be treated as a default condition of operation rather than an optional feature. This approach is particularly relevant in the context of microscopic drones, where invisibility undermines the basic assumptions of transparency and consent that underpin conventional data protection regimes. By integrating privacy safeguards at the level of system design, PbD attempts to compensate for the inability of individuals to perceive or respond to data collection in real time. In practical terms, this involves restricting data collection to clearly defined operational purposes and minimising the volume of personally identifiable information generated by the system. Techniques such as real-time anonymisation, selective sensor activation, and spatial or temporal limitations on data capture can reduce the extent to which micro-drones collect or retain sensitive information (Cavoukian, 2012, pp. 18–23; European Union Aviation Safety Agency, 2024). These measures reflect a shift toward functional proportionality, in which data collection is aligned with specific tasks rather than continuous or indiscriminate monitoring. However, their effectiveness depends on the integrity of implementation and the willingness of developers and operators to prioritise privacy over performance or convenience. A more fundamental challenge arises from the fact that many PbD principles presuppose some degree of system visibility or user awareness. In the case of microscopic drones, where platforms may be effectively undetectable, technical safeguards must substitute for the social and legal mechanisms that ordinarily enable individuals to exercise control over their data. This has led to proposals for hardware-based indicators, remote identification systems, and other forms of embedded transparency designed to make otherwise invisible systems at least partially observable (European Union Aviation Safety Agency, 2024). While such measures can improve traceability, they are difficult to implement consistently at very small scales and may conflict with operational requirements, particularly in security or military contexts where low observability is a defining feature of the technology.

From a governance perspective, PbD represents a partial but insufficient response to the challenges posed by micro-scale systems. It can mitigate some risks by reducing data exposure and improving accountability, but it does not address the underlying structural problem that these systems operate outside the conditions required for meaningful consent and oversight. In particular, PbD assumes that design choices can substitute for regulatory enforcement, an assumption that may not hold where incentives favour data extraction or covert operation. As a result, PbD should be understood not as a standalone solution, but as one component of a broader governance approach that combines technical safeguards with institutional and legal mechanisms capable of addressing the systemic implications of invisibility.

5.3 Responsible Research and Innovation (RRI) Pathways in Micro-Robotics

Responsible Research and Innovation (RRI) provides a complementary framework for addressing the dual-use and governance challenges associated with microscopic robotics by embedding ethical and societal considerations directly into the research and development process. Rather than treating innovation and risk as separate domains, RRI emphasises anticipation, reflexivity, and responsibility, requiring researchers to consider the broader implications of their work throughout the lifecycle of technological development (Drukarch et al., 2023, pp. 15–19). In the context of micro-scale systems, this approach is particularly relevant because the same characteristics that enable beneficial applications—such as medical intervention or environmental monitoring—also create opportunities for covert surveillance or weaponisation. As a result, managing dual-use risks cannot be deferred to regulators alone but must be addressed at the level of design, dissemination, and research practice. A central element of this approach is the recognition that openness in scientific research is not an absolute value, but a variable that must be balanced against potential risks. In fields such as robotics and artificial intelligence, where open-source practices accelerate innovation, unrestricted dissemination of designs, code, and datasets can also lower barriers to misuse (Righetti & Boulanin, 2026, pp. 5–6). RRI therefore encourages a graduated model of openness, in which access to particularly sensitive information is managed through mechanisms such as staged release, controlled repositories, or peer-based screening. These measures do not eliminate the possibility of misuse, but they can slow diffusion and create opportunities for risk assessment before technologies become widely accessible. At the same time, RRI highlights the importance of defining normative boundaries within the research community itself. The establishment of “red lines” for unacceptable applications—such as the deliberate design of systems for autonomous lethal use or covert mass surveillance—represents an attempt to articulate shared standards of responsibility (Righetti & Boulanin, 2026, p. 1). Such norms are inherently difficult to enforce, particularly in a globally distributed research environment, but they play a role in shaping expectations and guiding professional conduct. In the case of microscopic drones, where capabilities are often ambiguous and evolve rapidly, the existence of such norms can provide an initial layer of governance in the absence of clear legal frameworks.

Effective implementation of RRI also depends on broader institutional and organisational structures. This includes integrating ethical and dual-use considerations into engineering education, requiring risk assessments as part of research funding processes, and establishing mechanisms for interdisciplinary collaboration between technologists, policymakers, and end-users (Ilovača, 2025, p. 45; Drukarch et al., 2023, p. 12). These measures aim to ensure that responsibility is not treated as an external constraint, but as an intrinsic component of technological development. In particular, early engagement with regulators and stakeholders can help identify potential risks and governance gaps before systems are deployed at scale. However, like Privacy by Design, RRI has inherent limitations. Its effectiveness depends largely on voluntary compliance and shared norms, which may be unevenly applied across different actors and jurisdictions. In competitive or security-sensitive contexts, incentives may favour rapid development and deployment over precautionary restraint. Moreover, RRI does not resolve the structural challenges posed by technologies that are inherently difficult to detect, attribute, or control once deployed. It can shape the trajectory of innovation, but it cannot fully substitute for formal governance mechanisms capable of managing systemic risks. As a result, RRI should be understood as a necessary but insufficient component of a broader regulatory framework, contributing to risk mitigation at the level of research practice while leaving unresolved the deeper problem of governing invisible and distributed systems.

5.4 Bridging the Translation Gap: A Value-Centered Readiness Framework (mTRL)

The transition from laboratory innovation to real-world deployment in microscopic robotics is constrained by a persistent translation gap, often described as the “valley of death,” in which technically viable systems fail to achieve operational or commercial adoption. This gap is not solely a function of technical limitations, but reflects a broader misalignment between engineering priorities, regulatory requirements, and practical use conditions (Ceylan et al., 2025, pp. 4–6). In academic environments, incentives frequently favour novelty and proof-of-concept demonstrations over system integration, reliability, and compliance, resulting in prototypes that are optimised for publication rather than deployment. At the same time, developers often encounter uncertainty regarding regulatory classification, particularly where micro-scale systems do not fit clearly within existing categories such as medical devices, consumer electronics, or industrial tools (Drukarch et al., 2023, pp. 11–14). This combination of technical, institutional, and legal barriers limits the ability of promising technologies to move beyond controlled environments. The milli/microrobot Technology Readiness Level (mTRL) framework represents an attempt to address this gap by reorienting development toward practical application and societal value. Unlike traditional Technology Readiness Levels, which emphasise technical maturity, mTRL introduces a “value pull” logic in which success is defined by alignment with user needs, regulatory requirements, and operational contexts (Ceylan et al., 2025, pp. 7–8). This approach requires that considerations such as safety, usability, and system integration be incorporated at early stages of development, rather than treated as late-stage constraints. In the context of microscopic drones, this is particularly important because the risks associated with invisibility, autonomy, and distributed operation cannot be addressed solely through technical performance; they must be integrated into the design and evaluation process from the outset. A key feature of the mTRL framework is its emphasis on iterative development across both technical and contextual dimensions. Progression through the readiness levels depends not only on improvements in hardware and control systems, but also on demonstration of functionality in realistic environments and alignment with existing infrastructures (Ceylan et al., 2025, pp. 8–11). This dual focus highlights the importance of system-level integration, including compatibility with external control systems, data processing architectures, and regulatory procedures. For micro-scale systems, which often depend on hybrid architectures combining onboard capability with external support, this integration is a critical determinant of viability. Without it, advances in individual components do not translate into operational capability. The framework also underscores the importance of early engagement with stakeholders, including regulators, end-users, and domain specialists. Mechanisms such as the LIAISON model seek to create feedback loops between developers and policymakers, allowing regulatory considerations to inform design decisions and vice versa (Drukarch et al., 2023, pp. 6–8). This approach contrasts with traditional models in which regulation follows deployment, often in response to observed harms. By integrating governance considerations into the development process, mTRL aims to reduce uncertainty and facilitate smoother transitions from research to application.

However, the applicability of such frameworks to microscopic drones is constrained by the same factors that complicate their governance more broadly. Systems that are difficult to detect, attribute, or monitor challenge the assumptions on which regulatory validation depends. In addition, the diversity of applications—from medical nanorobots to distributed sensor networks and micro-scale aerial systems—limits the extent to which a single framework can provide comprehensive guidance. While mTRL improves alignment between technical development and practical requirements, it does not resolve the underlying tension between innovation and control. As with Privacy by Design and Responsible Research and Innovation, it contributes to mitigating specific aspects of the problem, but does not eliminate the structural governance challenges associated with technologies that operate below conventional thresholds of visibility and oversight.

6. Conclusion

The central argument of this study is that the miniaturisation of robotic systems constitutes a qualitative transformation in the relationship between technological capability and governance. As systems move into the sub-gram and millimetre domain, the reduction of physical scale does not simply diminish impact; it reconfigures the structure of risk. The defining feature of microscopic drones is their ability to operate below the thresholds of detection, attribution, and control on which existing regulatory and security frameworks depend. This “invisibility paradox” shifts the locus of concern away from kinetic effects toward systemic vulnerabilities, including surveillance, data extraction, and loss of accountability. In this sense, the most significant consequence of miniaturisation is not technical but strategic: it enables forms of action that are difficult to observe, attribute, or regulate. At the same time, the analysis demonstrates that the capabilities of microscopic drones remain tightly constrained by physical and engineering limitations. Persistent energy bottlenecks, aerodynamic instability, limited sensing and communication capacity, and the challenges of autonomous navigation impose a hard ceiling on performance. It is therefore essential to distinguish between demonstrated capabilities and projected applications. While laboratory systems have achieved controlled flight, distributed sensing, and limited autonomy, their effectiveness in open, real-world environments remains restricted. This gap between technological potential and operational reality is particularly important in assessing military relevance. Micro-scale systems do not currently provide a substitute for conventional platforms, nor do they enable decisive forms of force projection. Their significance lies elsewhere.

From a military and security perspective, the primary impact of microscopic drones is the expansion of operational space below the threshold of conventional warfare. Their small size, low cost, and potential for large-scale deployment enable new forms of distributed sensing, surveillance, and disruption that are difficult to detect and counter. These systems are particularly suited to tasks involving proximity and persistence, including intelligence gathering in denied environments, monitoring of critical infrastructure, and low-level harassment or interference. In this context, their value derives not from individual performance but from collective behaviour, redundancy, and low observability. This creates a class of capabilities that are inherently asymmetric, favouring actors able to exploit scale and concealment rather than technological superiority in the conventional sense. The risks associated with these developments are correspondingly systemic. The combination of invisibility and autonomy produces an attribution gap in which actions cannot be easily linked to responsible actors, complicating deterrence and accountability. In a military context, this raises the possibility of covert or deniable operations that blur the boundary between peace and conflict. Distributed micro-scale systems could be used to conduct persistent surveillance, disrupt communications, or degrade infrastructure without triggering clear thresholds for response. While such actions may not be strategically decisive, they have the potential to increase instability by enabling continuous, low-level interference that is difficult to attribute and therefore difficult to deter. This dynamic is consistent with a broader shift toward sub-conventional and hybrid forms of conflict, in which ambiguity and deniability are central features. At the same time, the vulnerabilities of these systems must be emphasised. Resource constraints limit their security, making them susceptible to interception, manipulation, and exploitation. Swarm-based architectures introduce additional risks, as compromised nodes can propagate errors or malicious inputs across the network. Environmental factors further constrain reliability, particularly for aerial systems operating in uncontrolled conditions. These limitations suggest that the near-term use of microscopic drones in military contexts will depend on integration within larger systems, rather than fully autonomous deployment. Their effectiveness will therefore be contingent on supporting infrastructures for communication, control, and data processing, which remain potential points of failure.

The analysis of governance frameworks indicates that existing approaches are insufficient to address these challenges. Regulatory systems based on visibility, physical classification, and identifiable actors are poorly suited to technologies that are inherently difficult to detect and attribute. While approaches such as Privacy by Design, Responsible Research and Innovation, and value-centered readiness frameworks can mitigate specific risks, they do not resolve the underlying structural problem. Effective governance requires a shift toward models that prioritise traceability, accountability, and outcome-based regulation, rather than reliance on physical characteristics of the platform. This implies the development of technical standards for identification, data logging, and system transparency, as well as institutional mechanisms capable of responding to distributed and ambiguous forms of activity. In conclusion, microscopic drones do not fundamentally alter the balance of military power in the near term, but they do change the conditions under which power is exercised. By enabling low-cost, low-visibility, and potentially deniable forms of action, they expand the range of operations below the threshold of conventional conflict, increasing the potential for persistent instability. Their long-term significance lies in their integration into broader technological systems, including artificial intelligence, distributed sensing networks, and cyber-physical infrastructures. Managing the risks associated with these developments will require not only technical innovation but a rethinking of governance, one that recognises that the primary challenge is no longer the control of visible platforms, but the regulation of invisible and distributed capabilities.

References

Anderson, R., Chan, H., & Perrig, A. (2003). Key Infection: Smart Trust for Smart Dust. Proceedings of the 11th IEEE International Conference on Network Protocols (ICNP’03), 206–215.

Bhushan, R., Sonker, P., Singh, P., Singh, N., Rai, S., Philip, R. R., Malik, S., & Dokal, T. C. (2026). Nanobots in medicine and beyond: From targeted drug delivery to intelligent environmental remediation. Next Materials, 11, 100396.

Cavoukian, A. (2012). Privacy and Drones: Unmanned Aerial Vehicles [Research report]. Information and Privacy Commissioner, Ontario, Canada.

Ceylan, H., Sinibaldi, E., Misra, S., Pasricha, P. J., & Hutmacher, D. W. (2025). Translating Milli/Microrobots with A Value-Centered Readiness Framework [Research report]. arXiv. https://arxiv.org/abs/2510.12090

Cheah, W. C., Watson, S. A., & Lennox, B. (2019). Limitations of wireless power transfer technologies for mobile robots. Wireless Power Transfer, 6, 175–189.

Chen, Y., Zhao, H., Mao, J., Chirarattananon, P., Helbling, E. F., Hyun, N. P., Clarke, D. R., & Wood, R. J. (2019). Controlled Flight of a Microrobot Powered by Soft Artificial Muscles. Nature, 575, Article 7782.

Coffey, T., & Montgomery, J. A. (2002). The Emergence of Mini UAVs for Military Applications. Defense Horizons, Article 22.

Computing Community Consortium. (2020). Assured Autonomy: Path Toward Living With Autonomous Systems We Can Trust (U. Topcu, N. Bliss, N. Cooke, M. Cummings, A. Llorens, H. Shrobe, & L. Zuck, Eds.) [Research report]. Computing Community Consortium.

Coppola, M., McGuire, K. N., De Wagter, C., & de Croon, G. C. H. E. (2020). A Survey on Swarming With Micro Air Vehicles: Fundamental Challenges and Constraints. Frontiers in Robotics and AI, 7, Article 18.

Davis, W. R., Jr., Kosicki, B. B., Boroson, D. M., & Kostishack, D. F. (1996). Micro Air Vehicles for Optical Surveillance. The Lincoln Laboratory Journal, 9, Article 2.

Drew, D. S. (2018). The Ionocraft: Flying Microrobots With No Moving Parts [Doctoral dissertation, University of California, Berkeley].

Drukarch, H., Calleja, C., & Fosch-Villaronga, E. (2023). An iterative regulatory process for robot governance. Data & Policy, 5, 1–22.

European Union Aviation Safety Agency. (2024). Privacy by Design [Research report]. European Union Aviation Safety Agency. https://www.easa.europa.eu/en/domains/civil-drones/privacy/privacy-by-design

Ho, Y., Tagliabue, A., Maliseti, S., Chen, Y., How, J. P., & Kong, K. (2024). Aerobatic maneuvers in insect-scale flapping-wing aerial robots via deep-learned robust tube model predictive control. Science Advances, 10, Article 19.

Hopster, J. K. G., & Maas, M. M. (2024). The technology triad: disruptive AI, regulatory gaps and value change. AI and Ethics, 4, 1051–1069.

Huber, A. F., II. (2002). Death by a Thousand Cuts: Micro-Air Vehicles in the Service of Air Force Missions (Occasional Paper 29). Air University.

Huckelberry, J., Zhang, Y., Sansone, A., Mickens, J., Beerel, P. A., & Reddi, V. J. (2024). TinyML Security: Exploring Vulnerabilities in Resource-Constrained Machine Learning Systems [Technical report]. arXiv. https://arxiv.org/abs/2411.07114

Ilovača, K. (2025). Implementation of Dual-Use Technologies in Defense and Public Security. The American Journal of Political Science Law and Criminology, 7, Article 06.

International Committee of the Red Cross. (2021). ICRC position on autonomous weapon systems [Position paper]. International Committee of the Red Cross. https://www.icrc.org/en/document/icrc-position-autonomous-weapon-systems

Jafferis, N. T., Helbling, E. F., Karpelson, M., & Wood, R. J. (2019). Untethered flight of an insect-sized flapping-wing microscale aerial vehicle. Nature, 570, Article 7762.

James, J., & Fuller, S. (2021). A high-voltage power electronics unit for flying insect robots that can modulate wing thrust. 2021 IEEE International Conference on Robotics and Automation (ICRA), 11373–11379.

Kahn, J. M., Katz, R. H., & Pister, K. S. J. (1999). Next Century Challenges: Mobile Networking for “Smart Dust.” Proceedings of the 5th Annual ACM/IEEE International Conference on Mobile Computing and Networking (MobiCom ’99), 271–278.

King, A. (2023). Robot Wars: Autonomous Drone Swarms and the Battlefield of the Future [Research report]. University of Exeter.

Kröger, J. (2019). Unexpected Inferences from Sensor Data: A Hidden Privacy Threat in the Internet of Things. In L. Strous & V. G. Cerf (Eds.), IFIPIoT 2018, IFIP AICT 548 (pp. 147–159). Springer. Li, Y., Zhu, M., Karnaushenko, D. D., Li, F., Qu, J., Wang, J., Zhang, P., Liu, L., Ravishankar, R., Bandari, V. K., Tang, H., Qu, Z., Zhu, F., Weng, Q., & Schmidt, O. G. (2023). Microbatteries with twin-Swiss-rolls redefine performance limits in the sub-square millimeter range. Nanoscale Horizons, 8, 127–132.

Lin, L., Sun, F. L., & Yang, W. Y. (2025). Untethered subcentimeter flying robots. Science Advances, 11, Article 5. Liu, S., & Chakkaravarthy, M. (2026). Lightweight security architecture for resource-constrained IoT devices: design patterns and implementation trade-offs. Future Digital Technologies and Artificial Intelligence, 02, Article 01.

Mankel, N. (2025, June 20). Chinese military robotics lab creates mosquito-sized microdrone for covert operations. South China Morning Post. https://www.scmp.com/news/china/science/article/3315206/chinese-military-robotics-lab-creates-mosquito-sized-microdrone-covert-operations

Microscopic Drone Technologies: Mapping Potential Benefits, Societal Risks, and Governance Gaps Across Insect-Scale Robots, Smart Dust, and Micro/Nanorobots. (2024). (Proposed Study Outline).

Mondal, I., & Haick, H. (2025). Smart Dust for Chemical Mapping. Advanced Materials, 37, Article 19.

O’Neill, P., Cranny-Evans, S., & Ashbridge, S. (2024). Assessing Autonomous Weapons as a Proliferation Risk: The Future Has Not Been Written [Occasional Paper]. Royal United Services Institute.

Pelletier, A., & Mueller, T. J. (2000). Low Reynolds Number Aerodynamics of Low-Aspect-Ratio, Thin/Flat/Cambered-Plate Wings. Journal of Aircraft, 37, Article 5.

Righetti, L., & Boulanin, V. (2026). Is open robotics innovation a threat to international peace and security? arXiv, 1–8. https://arxiv.org/abs/2601.10877v1

Sesham, P. K., Sesham, S., & Ndiaye, M. (2026). Ensemble learning models for micro-drone detection using integrated acoustic signatures. Discover Artificial Intelligence.

Shen, W., Peng, J., Ma, R., Wu, J., Li, J., Liu, Z., Leng, J., Yan, X., & Qi, M. (2024). Sunlight-powered sustained flight of an ultralight micro aerial vehicle. Nature, 631, Article 8021. Székely, J. (2024). Legal Aspects of Dual-Use Technologies: Emerging and Disruptive Technologies. In K. Zombory & J. E. Szilágyi (Eds.), Shielding Europe with the Common Security and Defence Policy. The EU Legal Framework for the Development of an Innovative European Defence Industry in Times of a Changing Global Security Environment (pp. 309–353). Central European Academic Publishing.

Wang, Y., Xia, H., Yao, Y., & Huang, Y. (2016). Flying Eyes and Hidden Controllers: A Qualitative Study of People’s Privacy Perceptions of Civilian Drones in The US. Proceedings on Privacy Enhancing Technologies, 2016, Article 3. Wu, B., Harvey, G., Whitbread, H., Li, H., Deng, D., & Xiao, J. (2022). Pushing the Limits of Micro battery Capacity and Energy [Research report]. Pacific Northwest National Laboratory.

Zając, M. (2025). Stopping Killer Robot Proliferation to Tyrants and Terrorists: Why A Global Ban Is Neither Necessary Nor Sufficient. Philosophy & Technology, 38, Article 86. Zhou, X., Wen, X., Wang, Z., Gao, Y., Li, H., Zhan, C., Huang, C., Fu, C., Zhu, S., Xiao, J., Luo, S., Wen, J., Shao, X., Xu, C., & Gao, F. (2022). Swarm of micro flying robots in the wild. Science Robotics, 7, Article 66.