Abstract
The development of UAVs with life-sign detection in the search-and-rescue (SAR) context challenges the assessment and mitigation of dual-use and misuse risks. Those technologies can be characterized as modularly constructed technologies (MCTs): They consist of highly specific components like sensors, UAV hardware, and often AI-supported software, and those components are designed in a way that makes them exchangeable and easy to implement. Therefore, the MCTs can be efficiently adapted for different purposes. From this situation, special dual-use challenges emerge. The recently finished project UAV-Rescue shall serve as an example for the development and assessment scenario of such an MCT from the SAR context. Referring to a recent paper from the context of autonomous driving, the contribution shows some mutual observations but goes further by (1) exploring the characteristics of MCTs that lead to special challenges in the assessment and mitigation of dual-use risks and (2) proposing a different way of dealing with these challenges. The central thesis is that MCTs cannot be addressed satisfactorily with a classic framework for dual-use classification and corresponding regulation. These are not just new types of technologies that also bear a dual-use risk among other risks, but a whole new type of highly complex technology that is designed to be adapted quickly and efficiently to different application scenarios. The paper argues that, even if it is difficult or impossible to mitigate those dual-use risks in MCTs with methods applied so far, it is highly important to provide a systematic analysis of the gains and losses that are caused by this technology. This is important to understand the irreversible impact of such developments in the sense of technology assessment on the one hand, and on the other hand to weigh costs against benefits in a circumspect and careful manner.
Keywords
Introduction
As technologies become more complex, assessing and mitigating dual-use and misuse risks becomes more challenging. A good example of this growing complexity and the increasing challenge of this assessment and mitigation are cases where high-performance sensors meet highly complex software, maybe even using artificial intelligence. Such a case can be found in rescue robotics with the development and practical application of unmanned aerial vehicles (UAVs) for search and rescue (SAR) operations. Originally coming from military research, UAVs (or drones) nowadays are attractive for a variety of civil application contexts, from logistics and mobility to research, art, documentation and leisure and benefit from previous military research and development of unmanned aircraft. As a result, enormous developments are now taking place in the civil domain that can vice-versa advance the technical possibilities also in the military domain. This applies very strongly to SAR drones that are capable of detecting people in unknown environments and provide a precise situation picture of the relevant environment. Those application scenarios are, for example, very similar to military remote sensing scenarios where people shall be traced, not for their benefit, but for purposes as catching, surveilling or even killing (Sandvik and Jumberg, 2018).
While civil and military development is being pursued in parallel, such a dual-use scenario is often to be considered problematic. Especially in rescue robotics there is a strong need for an ethical perspective to examine the ethical acceptability of research and development projects for responsible development, design and usage from the perspective of different stakeholders (Harbers et al., 2017). From this point of view, procedures of ethical evaluation and technology assessment that refer to civil usage obviously cannot be simply transferred to contexts that pursue a directly opposite purpose (not rescuing people but harming them); that would be function creep (Koops, 2021). But the challenge goes even further: If a highly dangerous and hard-to-fend-off weapon or surveillance system could be constructed on the basis of a humanitarian UAV system or even scenario-designated components of it, we must consider if we accept such a development, and if so, how it can be controlled so that it does not fall into the wrong hands, that is, what means could be used to mitigate such a risk based on a sound assessment of those risks.
My contribution argues that the previous means for reflection and regulation are difficult to apply in the case of modularly constructed technologies (hereafter referred to as MCTs) that contain both sensors and powerful software, even supported by AI technology and which are of great interest for various civilian contexts. My central point is that even if the dual-use risk is hard to mitigate in those cases, it is indispensable to evaluate the long-term acceptability of the whole type of technology under these aspects. This evaluation is crucial for long-term monitoring (rather than a one-time assessment) that creates a basis for regulation (Riebe, 2023) and also for societal awareness, which in turn has important potential for monitoring and participatory design of regulatory frameworks in the sense of responsible innovation (Genus and Stirling, 2018).
The case of human-detecting humanitarian UAVs
The use of UAVs (unmanned aerial vehicles, or drones) is quite familiar today in different civil contexts, although this technology comes originally from military research. Today, sensor-bearing UAVs, also with (semi-)autonomous navigation functions, are developed also for other than military purposes, for example drone photography, sports, logistics, mobility or humanitarian purposes. Historically, we know such ‘de-militarization’ (Boucher, 2015) of certain technologies from other examples, such as the intensive development of the internet or satellite technology. While these examples stand for peaceful use of technologies that can also serve as an infrastructure for military purposes, sensor-bearing drones could be turned directly into devices for killing, persecuting or surveillance. In the overall case of drones, that has led to a long-lasting debate on their dual-use potential, culminating in the question of the possibility of a ‘good drone’ (Sandvik and Jumberg, 2018).
Since the diffusion of drone technology on different markets is already quite advanced, and together with novel sensor technology an auspicious solution for difficult situations with vital importance like search and rescue missions (or contactless health monitoring in the case of a medical use), the case of sensor-bearing drones with functions related to remote sensing or even life-sign detection brings the dual-use debate to another level. With the development of drone technologies in very different contexts – SAR, remote sensing and weapon systems, even logistics and mobility –, the know-how of technological construction from basic research and from existing systems that can be rebuilt is broadly available for similar constructions. In detail, even if a life-sign detection device is explicitly constructed for a humanitarian purpose, the technical knowledge behind it enriches the pool of knowledge that is required to realize life-sign detection for purposes that are not beneficiary for the detected living being. While it might even be ethically required to apply state-of-the-art technology for humanitarian or medical purposes, we must ask if the development of a life-saving technology is acceptable even with an (obvious) dual-use risk. For a deeper understanding of both aspects – the growing complexity and the growing risk that is becoming increasingly difficult to manage –, it can be helpful to have a look at a recently finalized project in the current development of UAV technology for life-sign detection, also regarding the special requirements in the SAR context.
In the project UAV-Rescue: UAV-borne sensor technology for AI-based support of rescue missions, 1 multidisciplinary research was conducted on an AI-supported, sensor-carrying UAV designed to support emergency responders during rescue missions in semi-collapsed buildings. Combined with a ground station that consists of a remote control and a display device – such as a mobile tablet computer – to present the indoor map with additional information and configuration options, it represents a system that can help responders explore areas that are difficult to access. The project addressed disaster scenarios such as earthquakes or hydraulic heave which cause damage to buildings where survivors can be hidden without the possibility of drawing attention to themselves, for example after losing consciousness. Obviously, both responders and disaster victims benefit from this technology: They gain essential situational knowledge faster, and based on this information, they can decide whether and when to enter dangerous areas and then avoid hazards when entering them. The most significant technical innovation in this project was the combination of AI-supported vital sign detection via radar with lidar-based indoor mapping on a flying platform (Philippi et al., 2022; Stockel et al., 2024). This combination required sensor fusion with ego-motion compensation which was also a central focus in the project (Herschel et al., 2022).
Experts involved in this project came from the fields of AI development, indoor mapping, sensor technology, drone construction and sensor data fusion as well as from prospective end user organizations. Since the technology readiness level (TLR) was limited to a demonstrator for the basic functions and not to a market-ready product, the technical aim of the project was focused on the research concerning the integration of the components into a working system (which also included building a UAV from scratch since the indoor mapping functions and also a radar-based collision avoidance function were integrated into a semi-autonomous navigation unit). Nevertheless, the project was constantly evaluated and monitored from an ethical perspective to identify future potential risks arising from the usage of this novel technology and to help develop mitigation strategies.
This evaluation showed that the dual-use risk of life-detecting humanitarian UAVs emerges directly from the specific features of this novel technology, that is, a wide range of sensors (camera, thermal imaging, lidar, radar), highly developed software (data fusion, AI-supported data processing) and the generation of situational images for responders. It is intuitively clear that these technical features make this new technology relevant for dual-use purposes, like in similar cases of SAR drones (Cawthorne, 2024). The question, however, is to what extent they pose a dual-use or misuse risk, and which dual-use concept is appropriate here. Three of the dichotomies Rath et al. (2014) use to classify different dual-use concepts seem relevant here: civil versus military, benign versus malicious, peaceful versus non-peaceful. 2 While the first dichotomy, as mentioned above, dominates the drone discourse, the other two dual-use terms provide a better grasp of the ambivalent use of the UAV for life detection: The device can be used both for and against the person being tracked, that is, to save or to persecute them (in accordance with the scenario described by Cawthorne, 2024: 45). This is associated with a definition of dual use as technological ambivalence (Riebe, 2023: 20).
Of course, which dual-use or misuse scenarios are conceivable in this sense depends on the underlying security concept, which in turn is linked to a reference object that is to be protected from a threat by the relevant technology (Riebe, 2023: 20). If the reference object of the risk assessment is the detected person, it can be argued that the entire system was developed explicitly to rescue people and that using it to track people constitutes an abuse. However, this does not capture the full scope of the problem of MCTs, for which this drone serves as an example. Furthermore, the question of possible scenarios and actors can only be examined in the context of a more detailed consideration of MCTs.
The particularity of modularly constructed technologies
At first sight, this case may be similar to other cases where research and development lead to technologies that bring great scientific progress and benefits to humanity but can be used not only for peaceful purposes. In cases such as the usage of nuclear power or gain-of-function research, technological knowledge and tools are developed with which one can initiate and modify physical and biological processes with mostly irreversible consequences. The case of vital-sign detecting UAVs belongs to a fundamentally different type of dual-use relevant technologies: Here we are confronted with the – increasingly common – phenomenon that several high-performance technologies are combined, where both the components themselves and their combination have a wide range of peaceful applications as well as a high potential for non-peaceful use and misuse. This means that it is not sufficient to subject the individual components to a dual-use assessment; rather, their possible combinations and their technical preconditions must also be considered. So here we have the special case that, to put it colloquially, the whole is more than the sum of its parts.
In the following, I would like to describe those technologies as modularly constructed technologies (MCTs). I use this phrase by analogy with conceptualizations that describe phenomena similar to this case, but slightly different in their respective contexts, such as ‘complex products and systems (CoPS)’ (Hobday, 1998; Hobday et al., 2000), ‘reconfigurable manufacturing systems (RMS)’ (Feng et al., 2024) and, in a more concrete context, ‘multipurpose reconfigurable UAVs’ (Da Silva Ferreira et al., 2018). What these conceptualizations have in common is that they specify a certain kind of design principle. It is the design principle of modularity that has often been discussed in relation to software development, making applications versatile and scalable (e.g. Valipour et al., 2009). It is applicable but not limited to the case of UAVs (Kekec et al., 2013); a similar phenomenon has been described for the case of autonomous driving (Guntrum et al., 2023).
This phenomenon is so far only perceived and treated in specific contexts. But the meaning of the combination of multi-purpose components has taken on a new dimension today. This is due to new developments and to an accelerated pace in the field of hardware and software development that gives rise to a new situation (Guntrum et al., 2023). But while the aspects of increasing complexity and speed of research can still be represented under the title ‘new and emerging science and technologies’ (Swierstra and Rip, 2009), there is apparently no title that makes the phenomenon tangible in a broader sense for ethical discussions and dual-use assessment. To grasp this design principle in a less context-specific way, the term MCTs is supposed to characterize systems that are combined by components in the sense of modules. These modular components are designed and developed in a special way.
(1) The components are closed in themselves, which ensures trouble-free functioning. Such complex components are the result of highly specialized research.
(2) They are designed for multiple purposes which leads to a broad applicability. At the same time, they are often developed in specific contexts, like the sensors in the case of UAV-Rescue. In other words, they are multi-purpose by design.
(3) As they are not designed to stand alone, but always as possible modules of more complex systems, their success is measured by their high compatibility with different systems. This is ensured firstly by a strict division of labor within the system – the respective component fulfills exactly one function – and secondly by standardization (Guntrum et al., 2023: 67). Standardization refers to connections and interfaces as well as to methods of use and best practices for combinations. Therefore, by the clear regulation of input and output, those systems are highly interoperable.
The multi-purpose design of components that can be easily combined in different ways makes MCTs interesting for both military and civilian contexts. And MCTs can be found in both contexts. The phenomenon of MCTs in civilian and military development is part of a long-term development: It has been observed before that military technology is a complex product system (‘CoPS’), which in turn can be combined to form larger technical systems (Hobday, 1998). As Meunier and Bellais (2019) point out, especially drones are a case of larger technical systems that ‘are composed of multiple components hierarchically organized to produce an integrated functional system’ (Meunier and Bellais, 2019: 436). But the effects of this construction principle become also increasingly evident in the civilian sector (Feng et al., 2024). Together with new developments and the acceleration in the field of hardware and software development, driven forward by different stakeholders, the modular, multipurpose design leads to a ‘rather unique situation’ (Guntrum et al., 2023: 66). In this situation, also research is much more difficult to address than in the more typical dual-use relevant cases like, for example, gain-of-function research as separate field.
The aforementioned estimation that this is a novel situation comes from the context of autonomous driving, empirically examined by Guntrum et al. (2023). This context is technically very similar to that of drones, although it is discussed in quite different debates, particularly about dual-use relevance. So, the case discussed here can help to understand the ambivalence of MCTs regarding dual-use scenarios as well as the technical characteristics of the modular multi-purpose design. In this contribution, the authors use empirical studies to show that in the case of autonomous driving and in the context of autonomous systems in general, combinations of sensors and software are in use that challenge the dual-use assessment in a novel way. They identify various aspects that also can be observed in the context of humanitarian UAVs (or rescue robotics in general): First, they observe a broad applicability for the ‘numerous technologies that make AD [autonomous driving] possible’ (Guntrum et al., 2023: 66). Furthermore, those technologies consist of several components, like sensors or software, along with the knowledge of how to combine them. This knowledge implies an ‘overall systemic solution competence’ that is transferable in the sense of ‘building blocks that are used to solve domain-specific problems’, even if those technical solutions cannot be directly transferred to other areas of application and must be adapted, especially if they differ in the way of operation like air, ground, maritime or space (Guntrum et al., 2023: 66).
Secondly, they identify the problem of an increasing large variety of actors in research, development and application that manifests in fast parallel progress in the research and development of relevant components and at the same time their widespread availability and the broad interest they are arousing among various actors. While the research on autonomous systems takes place in different research fields and disciplines (Guntrum et al., 2023: 56), other types of actors, such as trade partners or private persons are involved, also in the role of target groups: ‘This ecosystem contains various and heterogenous actors such as automotive or robotic companies, universities, or the open source community’ (Guntrum et al., 2023: 57; Mallozzi et al., 2019). That comes with the fact that ‘technologies like unmanned aerial vehicles or small drones are increasingly globally accessible for non-state actors’ (Guntrum et al., 2023: 58; Dahlmann and Dickow, 2019).
Their résumé of both aspects, that is based on interviews with actors from different fields, is that the ‘collaboration of various technologies (e.g. AI, mechatronics) and different sector fields (e.g. aviation, automotive)’ and the combination of ‘different components of knowledge of heterogeneous systems’ make it ‘relatively easy to transfer technology modules into military solutions’ (Guntrum et al., 2023: 58; Meunier and Bellais, 2019). These circumstances go hand in hand with the phenomenon of a ‘fragmentation’ (Guntrum et al., 2023: 56) of the research field of autonomous systems that leads to an insufficient awareness to the dual-use risks emerging from the whole respective systems (Guntrum et al., 2023: 62) as well as it ‘hinders the clear identification of a specific dual-use technology’ (Guntrum et al., 2023: 53).
Considering the above characterization of MCTs from a developmental perspective, the following consequences for their implementation can be identified:
(1)
(2)
(3)
The special feature of MCTs is that one can use components that are already available and concentrate on combining them. Furthermore, the modular structure of the emerging technologies allows a fast and effective adaption of whole systems to other contexts. In the case of UAV Rescue, a central technical objective of the project is the data fusion of sensor and radar data and the development of reliable indoor mapping with SLAM and self-motion compensation, optimized for the respective application. The respective components and application scenarios are explored in several different contexts. For example, life-sign detection via radar is also interesting for mobility contexts, for instance to recognize pedestrians, and the measurement of vital signs can be a great benefit in medical applications (Kebe et al., 2020).
So, generally speaking, the dual-use risk of life-detecting humanitarian UAVs emerges directly from the specific features of this novel technology, such as a wide range of sensors (camera, thermal imaging, lidar, radar), highly developed software (data fusion, AI-supported data processing) and the generation of situational images for responders. In other words, the greatest humanitarian benefits of this technology create the strongest dual-use interests. On the one hand, those benefits are very specific to drone technology combined with sensors that enhance human environment perception (Choi-Fitzpatrick, 2014) and artificial intelligence, which enables efficient data processing as well as (semi-)autonomous operation. These include the features of (1) remote sensing in unknown indoor environments and (2) the rapid achievement of a large range. Both aspects are enormous advantages in the context of search and rescue operations and are also highly relevant in military contexts. The scenarios in which such a system can be helpful in the humanitarian and the military context are also quite similar, namely semi-collapsed buildings with unknown hazards and no other possibility to have a glance inside. 3
On the other hand, the development of (3) contactless vital sign detection using radar (Kebe et al., 2020) has created a detection method that is both imperceptible by the detected person as well as difficult to fend off due to the distance involved in the measurement. The fourth aspect that makes the SAR system and the insights gained during its development attractive for a dual-use scenario is (4) the asymmetry of perception and action associated with the system: The roles of the involved persons are strictly divided into active and passive, into detectors and detected. In the SAR context, the advantage is being able to see without putting oneself in danger. In the context of a weapon or surveillance system, this would result in the benefit of seeing without being seen and the possibility of doing harm without being in danger. In combination with aspect (3), the advantage of invisibility does not only apply to the human operator of the UAV system, but also to the measurement that can be carried out without the knowledge of the person being measured or detected. To conclude, the dual-use risks – depending on advantage (1) and (3) – consist of using the device to find somebody who is hiding (and does not want to be detected, in opposite to the humanitarian scenario) – even without the person noticing the vital sign detection or measurement, and depending on (1) and (2) just due to the similarity between SAR and military scenarios. 4 Especially aspect (4) brings an obvious, direct benefit for military, surveillance, and terrorist attack purposes.
The UAV demonstrator developed in the project is also an interesting example of the MCT principle in another respect. It shows that MCTs themselves can be the starting point for more complex systems: A sensor-carrying UAV like the one described above can be further enhanced by already developed technologies like thermal imaging with infrared. However, the case of drones is ambivalent. In the literature, drones are often considered to be a standard building block for more complex systems, for example as part of robot swarms (Grossman, 2018). In a simpler case, a drone capable of carrying loads and sensor technology such as cameras, radar or lidar systems, which are independent components that transmit information to a ground station where it is processed, is required to build such a UAV. In the exemplary project, a more complex system is developed that can navigate (semi-)autonomously. In this case, the control system must also be able to access sensor data. Therefore, the UAV itself must be constructed nearly from scratch, as was done in the UAV Rescue project. This requires a certain amount of cooperation, but there is a growing repertoire of resources: Such a system can also be constructed in a modular way with the appropriate ‘systemic solution competence’ (Guntrum et al., 2023), as it is possible to utilize construction components and even construction plans that are widely used in the hobby sector. This also applies to the necessary software, which may also be AI-supported.
So we are dealing here with a series of complex technical developments that are each already ambivalent in terms of their dual use and that are efficiently combined with each other, thereby adding up their potential:
(1) Sensor integration with AI-driven data processing: It enables 3D mapping with object detection (for its dual-use relevance cf. Grossman, 2018: 42 ff.).
(2) A sensor-bearing drone: As Choi-Fitzpatrick (2014) describes, the usage of sensor-bearing drones brings a shift not only of perception but especially of perspective since areas can be seen that would been hidden from the ground perspective (Choi-Fitzpatrick, 2014: 20). These devices offer increased range and sight, while the operator can also benefit from the fact that she is invisible.
(3) (Semi-)Autonomous navigation: The benefits of autonomous systems, in general, are ‘range, endurance, speed, and coordination, in comparison to existing non-autonomous piloted systems’ (Guntrum et al., 2023: 54; Scharre, 2018) – and this is even enhancing the existing advantages of remotely piloted systems.
In the study of Guntrum et al. (2023), the authors’ assessment that no clear dual-use technology can be identified requires further clarification. While they identify environmental perception, artificial intelligence and sensors as most relevant for knowledge transfer from civil research to the military sector, their suggestion for an adapted dual-use assessment is to pay more attention to sensor development. As shown above, of course, for a dual-use assessment that does justice to new and emerging technologies, it is crucial that more attention is paid to developments in individual areas such as sensor technology or artificial intelligence (Schmid et al., 2022), which supply components for highly complex combined systems. On the other hand, considering the components is not enough because, as outlined above, the design principle of MCTs has the effect that the whole is more than the sum of its parts.
This means that a completely new situation is emerging, which also requires a new type of dual-use assessment. Before frameworks or regulations can be provided for this, it is necessary to understand what exactly has changed with MCTs regarding previous dual-use relevant technologies and previous ways of dual-use assessment. This situation requires descriptions of this change as well as the identification and articulation of new aspects and the establishment of new categories of understanding. In order to tentatively approach an appropriate description of MCTs from the perspective of dual use, I would like to present three views on this complex phenomenon from interdisciplinary discourse. On closer inspection, they turn out to be spurious arguments, but they still help to grasp and describe the new situation because they reveal the friction between its demands and previous assessment methods. Each of them highlights a form of change from which a criterion for a successful dual-use assessment of MCTs can be derived. After that I will briefly compare these criteria with previous assessment methods in order to prepare the conclusion in the last section of my contribution.
How to deal with it?
Some spurious arguments
In the following, I will refer to some observations that do not belong to the scientific discourse in the context of dual-use research, but are opinions that can be heard in more informal situations in interdisciplinary contexts such as project consortia or spontaneous discussion rounds. These opinions often do not take the form of scientifically sound arguments, but rather that of prejudices. However, as Gadamer emphasized in his philosophy of hermeneutics, prejudices have an important orientational function (Gadamer and Weinsheimer, 2011). They usually contain a prior understanding of the relevant problems from specific perspectives that can be utilized for new descriptions of categories of understanding. An appropriate assessment of novel technologies is based on a broadly informed, conceptual pre-understanding of these technologies and their practical application to provide orientation in the problem formulation, from which concrete research questions can then be posed (Grunwald et al., 2023). Such a pre-understanding can arise from pre-opinions of different perspectives with their specific knowledge, priorities and caveats.
At the same time, those prejudices can also have undesirable effects, because they can shape the discourse and thus also hamper questions and, above all, the willingness to engage in dialog (Philippi, 2023). This poses a problem for dual-use awareness (Schwartz et al., 2022) and a willingness to accept and face this challenge. When discussing the risks of dual use and misuse of humanitarian UAVs like the one examined in UAV-Rescue in such an informal interdisciplinary setting, one often encounters various reactions that suggest or even explicitly assert the futility of a dual-use assessment for this context. I will briefly outline three spurious arguments in the following and evaluate them from the perspective of an appropriate dual-use assessment for MCTs. They can be clustered into three categories:
(1) ‘They already have it’. This spurious argument is aimed at the futility of a dual-use assessment of humanitarian drones: Military research is already at least as far along, it does not depend on results from civilian research. Guntrum et al. (2023) help in three ways to invalidate it in the case of autonomous driving technologies. First, they show that this claim is simply not correct. They do so with an observation that can be transferred considering the rapid progress in the areas of sensor technology and software in general: ‘With the rapid improvement of [autonomous systems], the military has a profound interest to implement civilian R&D into their products or technologies’ (Hansen and Sauer, 2019). ‘[. . .] the flow of information between civilian and military spheres has been reversed, as until the 1960s military research and knowledge mostly influenced civilian research.’ (Guntrum et al., 2023: 58; Kavouras and Charitidis, 2020) The use of the word ‘flow’ here suggests that this is not a one-time adoption, but rather that spillover is now a continuous recourse to technologies developed and widely available in the civilian sector. Secondly, it is not possible to check which collaborations already exist because a lot is classified (Guntrum et al., 2023: 69). Therefore, it remains largely speculation and cannot serve as a starting point for a proper dual-use assessment (or even a discard of it). And thirdly, the fact that they are not dependent on them does not mean that it is not attractive to benefit from these developments in terms of lower costs (Guntrum et al., 2023: 66). Furthermore, this spurious argument does not apply in cases involving non-state actors. But also in regard to the military case, it does not address the particular potential of developing dual-use relevant MCTs, that is to provide components and design knowledge that are also of interest for military research. So, the change that becomes visible here and must be considered in a proper assessment of MCTs, concerns the innovation transfer, which is increasingly flowing in both directions with MCTs; the decisive progress in the respective components, the combined knowledge and the proven combinations are increasingly taking the form of widespread state-of-the-art technology.
(2) ‘You can use a pencil for writing a military strategy. Or use a mobile phone to construct a bomb’. This argument which I would like to call the pencil argument is quite clearly a killer argument, stating that everyday technology can also be used for non-peaceful purposes. Especially the pencil example shows that this spurious argument aims at inflating and trivializing the concept of dual use regarding very basic cultural assets. Basically, it gets the character of those cultural assets wrong: If one uses the pencil for writing there is no dual use or misuse at all but a proper usage, independent from what is written with it. Nevertheless, this spurious argument touches an important point also regarding MCTs, raising the question of a proper dual-use assessment of broadly available technology that can be misappropriated.
It must be noted here that the mobile that serves as an igniter could be exchanged by other means. It is neither designated to be used for reaching the aim of a bomb ignition, nor by design necessary for it. Houkes (2008) refers to this type of application, which also uses new combinations of existing technologies, as ‘creative use’ (Cawthorne, 2024: 28) and Forge (2010) addresses this problem with his concept of improvised weapons (Riebe, 2023: 20). In this case, broadly available technologies are used for an alienated purpose in concrete situations. This links the respective risk assessment to specific scenarios and to current contexts: According to Forge (2010), the risk of misuse of technologies and thus their classification as dual-use relevant depends on current situations. His examples (car bombs in Northern Ireland, fertilizer bombs in Australia and Saudi Arabia) focus on the practices of certain groups at certain times to improvise weapons. To sum it up: In the case of ‘creative use’, technologies are alienated from their original purpose, while MCTs regularly consist of multi-purpose building blocks and combinational knowledge. The technique of improvisation is then replaced by the regular technique of combination of what is at one’s disposal to achieve their goal.
Of course, it is possible that a very complex device is designed for very specific and demanding application scenarios but can also be adapted relatively easily to similar ones. Misappropriating a drone built for humanitarian purposes for the opposite purpose is clearly an abuse and can be addressed with strategies such as capability caution (Cawthorne and Devos, 2020). But recombining parts of a drone developed according to the MCT principle for a different purpose is a proper use of the components, that is combining them into something useful, even if the resulting device is not in the spirit of the development of the drone itself that was advanced under the premise of peaceful use. The individual parts are indeed designed to be used for different purposes (unlike the mobile phone igniter); but sometimes it is the combination of the components that results in a dual-use case (such as radar-based life sign detection, especially on a flying platform). Nevertheless, it is the components, their design according to the MCT principle and ultimately a proven combination (such as a life-signs-detecting drone) that give rise to this disposition for non-peaceful and malevolent use. As Guntrum et al. (2023) point out in relation to autonomous (driving) systems, the combinability of continuously developed components and software not only increases the application range of these technologies but changes the whole situation fundamentally: ‘In retrospect, technologies could always be used for multiple purposes. But the progress to extend the capabilities of AS [autonomous systems] in combination with IT-related progress is a rather unique situation’ (Guntrum et al., 2023: 66). Therefore, the case of MCTs and their components is fundamentally different from the case of the pencil or the mobile phone as a bomb part, even if drones, sensor technology and AI-based software somedays similarly will be part of our everyday life like pencils or mobiles. Again, the change that is becoming visible here and must be taken into account in an appropriate assessment of MCTs concerns the assessment of potential misuse of them and their components in terms of misappropriation. In the case of MCTs, the criterion of misappropriation does not apply in the traditional sense to dual-use and misuse scenarios.
(3) ‘Mitigation is impossible, so talking about it is useless’. This spurious argument may follow from the above and is based on a kind of ‘reversed solutionism’ 5 : It tacitly presupposes that a problem that cannot be solved cannot be a problem. Identifying this one as a killer argument, based on a strong and problematic premise, should be sufficient for its response. At the same time, however, it can be seen as a sign of resignation, arising from the fact that previous methods of dual-use assessment and, in particular, mitigation strategies in the case of MCTs are no longer sufficiently effective. So finally, the change that is becoming visible here and must be taken into account in an appropriate assessment of MCTs concerns the new situation as a whole. Resignation is not a solution; the new difficulties do not release us from the responsibility to develop new standards and to promote dual-use awareness in the sense of Schwartz et al. (2022).
I have listed these often-heard spurious arguments to highlight that a responsible approach to complex technologies – and therefore also a thorough dual-use assessment – requires openness and creativity and the willingness to deal with a special form of ambivalence coming with the multi-purpose characteristics of MCTs, creating a fundamentally novel dimension of dual-use and misuse potential of technologies that are explicitly not designed to harm people. The following will show how this new dimension challenges the previous ways of classification and mitigation of dual-use risks.
Mitigation strategies and their limitations in the case of MCTs
To substantiate this diagnosis, four common strategies for minimizing dual-use risks are named below and applied to the case of MCTs.
(1) Mitigation of dual-use risks by design. This strategy can be applied within a development project. One example is shown in Cawthorne and Devos (2020), where the principle of capability caution is applied to a healthcare drone. It aims to design the developed system in such a way that it is as unattractive as possible for conceivable dual-use or misuse scenarios, for instance by addressing existing regulatory criteria against the background of ethical principles (Cawthorne and Devos, 2020: 1575; Novitzky et al., 2018: 80–82). Although this is an important step in addressing the dual-use risks in products designed as monolithic (i.e. non-modular) units, it is obvious that it cannot overcome the specific risks of an MCT and the components adapted to it. There still remains the possibility that the system can be rearranged with the same or similar components and corresponding adaptations for other contexts; and published research results on components and combination strategies can be used for further combinations such as sensor integration, especially since there is a blurring line between basic research and specific adaption to the combination of components.
(2) Export control. Export control is ‘a trade instrument of upholding international security objectives in the framework of non-proliferation of weapons of mass destruction’ (Alavi and Khamichonak, 2017: 59). In its present form, it is based on clear definitions of technologies and corresponding risks, as well as on at least a basic clarity of goals and actors. At the center of its regulations are control lists. It is precisely these fundamentals that are being complicated by the characteristics of emerging technologies (Rotolo et al., 2015), such as their difficulty in being defined, the wide range of purposes and actors involved, and their increasing importance for items that are more difficult to control, such as data and knowledge (Brockmann, 2018; Jones, 2021; Shagina, 2023). As Kavouras and Charitidis (2020: 197) state: ‘A legal framework composed of “static” regulations about DU will soon become impractical, if not obsolete, as new technologies arise’. For emerging technologies and especially their special case of MCTs, the definition of dual-use goods as ‘advance technological items that are not freely available on the global market and may be obtained from a limited number of exporters’ (Alavi and Khamichonak, 2017: 60) indeed becomes obsolete. Inside this EU-wide regulation system, especially dual-use relevant goods are listed as critical items and underlie strict control of export. The relation between complex systems and their components is constituted in a way that a system containing listed components underlies the same export regulations as the single component (e.g. radar sensors). In the case of MCTs, this will not be sufficient insofar as the single hardware components are available in civilian and international markets (e.g. radar and lidar sensor systems for automotive context). Regarding software and system solution competence, there are regulations affecting knowledge transfer, too, but they explicitly do not include basic research, 6 which is, however, crucial for the development of MCTs.
(3) Regulation of usage. This strategy involves restricting or even banning certain uses of certain technologies. It could be applied to the usage of particular MCTs like drones (Philpott et al., 2018; Storr and Storr, 2018) or certain hardware or also software components like AI, which is currently regulated by the EU AI Act (Regulation, 2024/1689). Highly relevant for the case of AI-driven MCTs is the current debate on banning autonomous weapons systems (O’Connell, 2023), which would include restricting the use of autonomous systems for killing people. It should be noted that such a ban could avoid military usage – given strong international regulation – but would not be directly effective concerning criminal or even terrorist misuse (Grossman, 2018; Scharre, 2017). Furthermore, this strategy requires a significant degree of communication between the various stakeholders involved and may mean foregoing options for action and development.
(4) Restriction of research and development. This is a very consequential strategy, especially in those cases where projects are indeed stopped, for example, in life sciences, but also in cases of other restrictions that influence the formation of knowledge in scientific communities (Wagner and Herington, 2021). However, there is a fundamental difference between research in application-oriented collaborative projects such as UAV-Rescue, and, for example, in the life sciences. There, too, digitalization is bringing about change, increasing both the pace of development and the dissemination of results, and producing tools that can potentially be misused (Heinrichs and Ergin Aslan, 2024). But whereas in the life sciences the dual-use relevant technology usually is the central object of research, research in application-oriented cooperation projects can remain basic research due to the explicit division of labor. This research is not initially classified as dual-use relevant, but its results are then integrated into the overall project, which may be dual-use relevant, in a further step. However, this is not the only problem if one wants to address dual-use relevant MCTs by restricting research because the development and testing of relevant components and their integration options also take place outside of science, namely in the commercial sector or – at a lower level – even in the leisure time of hobbyists.
As seen above, the dual-use potential of MCTs cannot be addressed inside the scope of a singular development project. Moreover, the mitigation of risks cannot be achieved by restricting knowledge transfer or export control but only by regulating which devices are allowed to be built for certain purposes. So, strategies (1) and (2) are not capable of mitigating the dual-use and misuse risks of MCTs sufficiently. Their problem is that they focus on certain technologies and their risks, not on combinations of components and software that can be attractive for a large variety of actors. This situation can only be addressed by strategies (3) and (4), requiring a lot of consideration and having a large impact on possibilities of action and development. It should be noted that the regulation of a component can lead to the regulation of the technology or its use. How efficient this is depends on whether and in what form the component is essential for the overall system. In the case of sensors, this is linked to a certain performance and can be easily circumvented by the principle of interchangeability; if artificial intelligence is involved, regulations of AI in the form of the EU AI Act could be helpful to at least assess the use of UAVs to detect vital signs as legitimate or illegitimate (Rath et al., 2014). However, such a regulation would not change the fundamental accessibility of the technology – as a set of components and combination knowledge – for various actors and scenarios. Therefore, it would be essential here to address an entire type of technology in the dual-use assessment, such as those that use AI (Schmid et al., 2022).
Why and how to talk about mitigation strategies for MCT dual-use nevertheless
To present here some of the requirements of a mitigation strategy for MCT dual-use and misuse risks can only remain an outlook. The first step toward a responsible treatment of how these new technologies affect research and development consists of acknowledging them and at the same time accepting the need for a novel perspective on the assessment of those technologies that are capable of dealing with their ambivalence mentioned above. The novelty lies above all in the possibilities that a dual-use assessment has to deal with this ambivalence and the extent to which the MCT design principle challenges previous methods and standards. While Guntrum et al. (2023) state correctly that the ‘fragmentation’ of the research on autonomous systems ‘hinders the clear identification of a specific dual-use technology’ (Guntrum et al., 2023: 53), this is primarily a problem if one wants to apply the previous categories. In this regard, dual use takes place in a much broader context than bound to certain technologies; it is about how we see and shape our world. To deal properly with the – in these regards – insufficient mitigation options by strategies (1) and (2), the perspective has to be widened.
The framework provided by Tucker (2012) for the case of emerging biological and chemical technologies (pp. 67–83) can help pave the methodic way. At the center of this framework is a flowchart for evaluating emerging dual-use technologies (Tucker, 2012: 69). In this flowchart, the risk of misuse is first assessed using the parameters of accessibility, ease of misuse, magnitude of potential harm, and imminence of potential misuse. The assessment is made at the levels of high, medium and low. In the case of a medium or high risk of misuse, the next step is to assess governability, depending on the more complicated parameters of embodiment (i.e. the tangible or intangible form of the technology as hardware, software or information), maturity, convergence (i.e. the number of disciplines involved in the development), rate of advance and international diffusion. Depending on the assessment of governability, hard-law, soft-law and informal measures are proposed. Only in the case of high and medium governability, a cost-benefit analysis is carried out and a tailored package of governance measures provided. Low governability returns the decision flow to the area of monitoring for possible changes, which then restart the assessment of risk of misuse and governability (as in the case of low risk of misuse).
Applied to MCTs, the framework shows an interesting picture. In their case, the parameters of accessibility and ease of misuse are typically estimated to be high, while the magnitude of potential harm and the imminence of potential misuse need to be assessed on a case-by-case basis. However, for the reasons given above, the governability tends to be low. In view of the considerations set out here, it is therefore possible that MCTs bear a high risk of misuse but remain within the scope of monitoring. It is important to recognize this situation (in contrast to spurious argument number 3). For the specific dual-use assessment of MCTs, I therefore would like to propose two promising amendments. They concern technology assessment and ethical evaluation as well as taking the challenge of dual use into account. They are aimed at long-term monitoring of such technical (and associated social) developments and therefore strongly address new and emerging technologies. Moreover, they should have the form of more flexible, case-by-case considerations that are better suited to such technologies than the existing frameworks (which is not to say that the development of suitable frameworks for MCTs is not useful and necessary).
The first change that needs to be made affects the form of the assessment. In light of Tucker’s decision flowchart (Tucker, 2012: 69), it is important to switch from the paradigm of a one-time assessment with a definite result to a steady monitoring and in-depth weighing of the risks and benefits of the respective MCTs. Due to the high complexity of these technologies and their application contexts, this will most likely have to be done on a case-by-case basis. Obviously, the previous parameters alone do not do justice to this new type of technology. Applying them to MCTs can lead to the resignation that is reflected in the killer arguments mentioned above. Above all, the monitoring still responds to changes that trigger a re-evaluation according to the mentioned parameters, but the meaning of that has to be clarified. On the one hand, this does not appear practical for a type of technology that is characterized by rapid and diverse change. On the other hand, the monitoring phase should be used for more than just the re-trigger, namely to develop conceptual knowledge about the new type of technology by observing changes, which may lead to an adaptation of the evaluation parameters and the possible strategies. The requirement for this is the acknowledgment of MCTs as a special case of emerging technologies and their intensive research as a new type.
The second amendment I would like to suggest concerns the perspective of the assessment, proposing a purposeful mix of perspectives. As seen above, a long-term monitoring is required for a proper weighing of benefits against risks, and this is regarding the legitimacy of research and development as well as the usage of a MCT in the very sense of strategies (3) and (4) above. Technology assessment and ethical evaluation are proper addresses for this concern in so far as they do not only deal with the impact and risks of technologies when used for their intended purpose but also with misuse and dual-use potentials and scenarios (Riebe, 2023). Furthermore, this applies not only for a one-time assessment of novel technologies but also for long-time monitoring and possible futures arising from new and emerging technologies. In the case of MCTs, the assessment must take place with the inclusion of interdisciplinary perspectives, which requires a high level of dual-use awareness (Schwartz et al., 2022) as well as a willingness to overcome disciplinary boundaries. Nevertheless, for example, in the case of rescue robotics, the awareness of dual-use risks is not always included in the ethical scope even if there is a broad consideration of risks and consequences (e.g. Battistuzzi et al., 2021), and as described above, it is not always an acknowledged topic in conducting interdisciplinary research projects. But due to the high convergence parameter, the situation for the assessment is different than for example in the case of gain-of-function research in the life sciences. In terms of an inclusion of dual-use and misuse risk reflection in technology assessment and ethical evaluation, this means reflecting not only on undesired secondary effects in the scope of its intended application, but considering a second, maybe harmful purpose (while ‘harmful’ must be differentiated in the respective case – such as use by non-democratic systems, against human rights, but in the case of SAR drones clearly against the interest of the detected person as described in Cawthorne, 2024).
Conclusion: Humanitarian UAVs as a visibility technology
My contribution was about the novel challenge that MCTs, such as life-detecting humanitarian UAVs, pose to a dual-use assessment as a novel type of technology. In particular, I wanted to emphasize that, despite the spurious arguments that are often heard, there is a dual-use relevance here that can and must be addressed. This is a first step toward a further dual-use assessment of life-detecting humanitarian UAVs. To conclude: A technology that is intentionally designed for applications in different contexts, as it is in the case of sensors and software, of course, has the potential for dual use and misuse. This applies even more in the context of MCT which represents a whole new cluster of technologies by combination of those components. The explicit aim of the multi-purpose design regarding implementation and compatibility and their relatively easy (re-)arrangement lead to technological items and skills that can be relatively easily adapted to other purposes by exchanging or extending several components. This is the reason why they pose new challenges for the assessment as well as the mitigation of dual-use risks. As shown above, there is a particular ambivalence here: On the one hand, the components are further developed in specific applications. However, they are designed for a variety of applications. And they are also proven in certain application scenarios, which – as in the case of the humanitarian use of vital-sign detecting UAVs – are dual-use relevant themselves due to their similarity to tracking or combat scenarios. The design principle of MCTs – creating broad combinability as well as accessibility – not only makes it more difficult to identify a specific dual-use technology (Guntrum et al., 2023) but also increases the range of actors and possible misuse scenarios, which in turn makes it more difficult to assess and thus mitigate risks emerging from MCTs and their components.
To give an idea of what this more specific monitoring might look like, I would like to conclude by focusing on a special type of MCT, which includes UAVs. In addition to the modularity aspect, Guntrum et al. (2023) not only highlight the novelty of the situation (‘In retrospect, technologies could always be used for multiple purposes. But the progress to extend the capabilities of [autonomous systems] in combination with IT-related progress is a rather unique situation’, Guntrum et al., 2023: 66) but also the peculiarity of technologies capable of environmental perception that ‘in particular stands out, as it represents a combination of software and hardware on the one hand, and a general application in [autonomous systems] with broad applicability on the other’ (Guntrum et al., 2023: 66) In this sense, a central dual-use aspect emerges from the fact that ‘software in combination with sensors is the main driver and a key aspect for functioning [autonomous systems]’ (Guntrum et al., 2023: 58). In the case of MCTs like in the project UAV-Rescue, environmental perception is not only a requirement of proper functioning systems like in the case of autonomous driving, but also the aim of the system itself. I would like to call this type of MCTs visibility technologies. They enhance vision in a very specific sense: They make visible what would otherwise remain invisible to us and in many situations create asymmetries in the freedom of action and thus power.
The ethical problem is in our example that the life-detecting UAV technology has a decisive impact on humankind: The loss of the possibility to be invisible, to hide, and, in the case of contactless vital sign measurement, to decide about informational self-determination in this regard. From this point of view, the specific dual-use scenario of transforming a humanitarian SAR device into a weapon system is not the only problem. 7 In general, a fundamental risk lies in the enhancement of human sight through obstacles and over large distances in a very precise way. These are the characteristics of this specific type of visibility technology, and those are necessarily MCTs. Even if this can only remain a rough sketch, it can be stated that against the background of the characteristics of such MCTs explained here, a sustainable dual-use assessment and management cannot avoid weighing up the benefits and costs in a very broad sense. To weigh risks against benefits means here to consider not only what is gained and even ethically required in concrete cases – the possibility of saving human lives with the current state of technology – but also what is lost in a long range. This results in a new dimension of technological ambivalence, which can be revealed by dual-use assessment, but then requires new categories of thinking. So, some theoretical development is required to keep pace with the novel ways of technological progress.
The crucial question that arises from this phenomenon is: We have to decide in which world we would like to live. This ‘we’ is not trivial, even if it sounds that way. The ‘we’ of military research or export control is not the ‘we’ of global technological progress as we currently experience it in the case of artificial intelligence or technologies based on it. To consider MCTs in this regard means to consider not a single technology but the impact of a wider technological development. Discourses of this kind are currently taking place intensively in the area of the use, risks and possible regulation of artificial intelligence. In the mid-20th century, they were held on the use of nuclear energy for peaceful and military purposes. But the logic of these discourses is also not entirely appropriate for the problem we face with MCTs such as life-sign detecting UAVs.
It is not just about more awareness of the previous type of dual-use awareness. Rather, it is about expanding this awareness to include new perspectives and technological potential and anchoring it in a broader discourse. Responding to the particularities described above may involve recognizing the potentially lost opportunities as values worth protecting. In turn, protection is conceivable by safeguarding these values through regulation or by developing technical means to preserve them. But this needs to be seen in order to be developed. In line with Riebe’s (2023) statement that risk assessment in dual-use assessment depends on a reference object that needs to be protected, one can say here it is the human ability to protect themselves by hiding from prosecution and killing. A dual-use assessment that thinks in terms of a modular and thus highly efficient and applicable technology development, such as that which produces visibility technologies, then addresses not only the ‘we’ of national security, but also of global humanity, which is affected by technology development. This distinction is already common in dual-use evaluation (Riebe, 2023: 22). Sensitized to technologies that arise from the design principle of the MCT, it can be a sustainable way to ensure responsible development, use and, if necessary, an informed regulation and restriction of MCTs and their multi-purpose components.
Footnotes
Acknowledgements
I would like to thank the participants of the BMBF summer school “Dual Use and Misuse of Research Result” for fruitful discussions and two anonymous reviewers for their valuable comments.
Ethical considerations
Non applicable.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research and authorship were supported by the German Federal Ministry of Education and Research under Grant 13N15549.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.