Autonomous Systems and the Speed of Battle: Legal Risks and Strategic Adaptation in AI-Enabled Warfare

Tactical Innovations in UxS Employment

Modern combat UxS are equipped with electro-optical and thermal imaging sensors that enable real-time reconnaissance and target designation across diverse operational and environmental settings, including low visibility, adverse weather, and contested electromagnetic conditions (Shen et al., 2025). Integrated machine learning algorithms enable these systems to autonomously detect and prioritize targets and coordinate actions within a group without requiring continuous operator input. The application of deep learning techniques, in particular, enables the recognition of complex movement patterns in high-risk environments. However, the underlying models rely on probabilistic classification methods and are currently unable to distinguish civilians from combatants reliably. Dong et al. (2023) report an accuracy rate of 91.2%, which remains unacceptably low in the context of armed conflict, where a single misclassification can result in serious violations of international humanitarian law.

The future integration of neural networks may enhance UxS’s capacity to adjust routes autonomously, select targets, and refine tactical behavior in response to changing battlefield conditions (Thomas, 2024; Xu et al., 2020; Salzmann et al., 2022). However, these functionalities remain experimental and raise critical concerns about human operators’ ability to retain meaningful judgment when machine decisions are executed in near-millisecond timeframes. Any prospective gains in operational efficiency must be carefully weighed against the risks of reduced human oversight, unpredictable algorithmic behavior, and potential legal breaches arising from inaccurate identification (Salzmann et al., 2022).

The adoption of modular hardware and software architectures enables the integration of diverse UxS into unified tactical networks (Hashim, 2024). This configuration allows small units to coordinate the simultaneous deployment of multiple platforms, facilitating concentrated precision strikes or synchronized reconnaissance operations. The collective interaction in swarm formations increases platform survivability and reduces response times to emerging threats to just a few minutes. An additional operational advantage lies in the capacity for rapid reconfiguration: both hardware (such as sensor arrays, communication modules, and combat payloads) and software components (including machine learning algorithms tailored to specific mission profiles) can be replaced to meet evolving tactical demands (Campion et al., 2019). The use of decentralized mesh networks further enhances operational robustness by allowing communication across distributed nodes and ensuring mission continuity even when contact with command centers is interrupted. However, UxS’s capacity to continue operating autonomously in such cases raises significant concerns about the preservation of meaningful human control and the attribution of legal responsibility when supervisory links are weakened or severed entirely (Mohsan et al., 2023; Campion et al., 2019).

Alongside the increasing operational capabilities of UxS, the development of countermeasures has also intensified. Efforts focus on the creation of interceptors (Wang et al., 2020), radio-frequency jamming systems (Wei et al., 2021), electronic warfare (EW) technologies (Yu et al., 2025), and directed-energy weapons designed to disable adversary-controlled platforms. High-power laser systems, for instance, can target and damage critical drone components, such as wings and sensors, at distances exceeding two kilometers, effectively neutralizing their control mechanisms. This technological interplay produces a continuous cycle of attack and defense within the unmanned domain of contemporary conflict. As one side advances new algorithms to evade EW interference, the opposing side simultaneously develops enhanced detection techniques and increasingly automated systems to launch countermeasures (Simpson, 2025) regarding Directed Energy Weapons.

Tactical innovations in UxS are transforming conventional approaches to reconnaissance, fire support, and maneuver warfare. The integration of swarm tactics with autonomous control algorithms enables the formation of adaptable combat groups capable of executing multiple reconnaissance-strike missions from various directions in parallel. A notable example is the experimental SkyGuard prototype, which illustrated how a coordinated swarm of UxS can conduct an “airspace saturation” maneuver over a designated area, facilitating the rapid identification and neutralization of both aerial and ground threats (Floreano & Wood, 2015). In parallel with these advancements, active interception capabilities continue to evolve. As UxS become more agile and autonomous, the need arises for responsive EW systems and AI tools capable of dynamically predicting drone movement and targeting patterns, thereby ensuring effective counteraction under high-speed operational conditions.

By 2030, the emergence of intelligent swarm systems is anticipated, with capabilities to autonomously determine tactics, assign targets, and coordinate actions without direct operator oversight (Kolling et al., 2016). Increasing attention is being given to air-ground hybrid swarms that integrate various types of UxS, including fixed-wing drones, hexacopters, ground robots, and uncrewed surface vessels. These heterogeneous groupings are expected to perform synchronized reconnaissance and strike operations with heightened efficiency. Concurrently, advancements in autonomous systems are focused on real-time adaptive learning. By applying federated learning and edge-AI technologies, these systems are designed to dynamically refine operational algorithms, enhancing responsiveness and functional resilience under highly volatile battlefield conditions (Konečný et al., 2016; Shi et al., 2016).

In parallel with the evolution of tactical UxS, several additional technological trends are expected to shape future capabilities. These include integrating quantum sensors and sixth-generation (6G) communication systems. Quantum interferometers embedded onboard may enable the detection of minimal ground or aerial disturbances, such as those generated by the movement of heavy armored vehicles. Simultaneously, 6G networks promise ultra-low latency (under 1 millisecond), facilitating real-time tactical data exchange among large-scale UxS groupings. The use of blockchain technology is projected to expand in securing communication protocols; decentralized ledgers will ensure the authentication of commands, thereby mitigating the risks of spoofing or unauthorized command alterations during missions. Furthermore, the development of energy-autonomous UxS is underway, focusing on high-efficiency graphene-based batteries and integrated solar panels, which are expected to support prolonged operational endurance in various environments (Rovira-Sugranes et al., 2022).

The synergy between modular hardware-software architectures, advanced machine learning algorithms, and state-of-the-art semiconductor technologies underpins tactical innovation in UxS. These combined elements ensure a high degree of adaptability, operational resilience, and the ability to operate at unprecedented speeds in dynamic, high-risk combat environments.

Uncrewed Water Systems and Defense

Uncrewed water systems (UWS) are increasingly embedded in contemporary naval doctrines, enabling coastal and maritime states to execute complex combat and support operations with minimal risk to personnel and optimized resource expenditure (Suman-Chauhan et al., 2024; Brock & Stone, 2024; Defense Security Monitor, 2024). Both uncrewed surface vehicles (USVs) and uncrewed underwater vehicles (UUVs) integrate high-precision sensor technologies with artificial intelligence algorithms, allowing for the autonomous execution of patrols, surveillance, threat detection, anti-submarine warfare (ASW), and mine countermeasure operations. These capabilities are particularly vital in technologically advanced theaters where adversaries employ sophisticated electronic warfare systems. Under such conditions, UWS must operate independently of disrupted global navigation satellite systems (GNSS) by switching to inertial navigation and employing adaptive control algorithms Harper (2024, October 28).

In the domain of intelligence, surveillance, and reconnaissance (ISR), USVs rely on a combination of GNSS, standardized inertial measurement units, and integrated multispectral sensors to enable autonomous navigation and course correction despite radio-frequency interference (Suman-Chauhan et al., 2024). Meanwhile, autonomous UUVs play a pivotal role in the underwater domain, serving as an initial detection layer. Equipped with both passive and active sonar arrays, they scan the acoustic environment to identify distinct submarine signatures, providing early warning data and refining targeting coordinates for integrated groups of autonomous platforms (Defense Security Monitor, 2024).

The capability for long-endurance missions is exemplified by UUV platforms of the Russian Cephalopod class, which can operate at depths of up to 6,000 meters and hold fixed positions for extended periods – a crucial feature for surveillance in strategically sensitive or maneuver-constrained areas (Drone Wars UK, 2024). Similarly, the Manta Ray program, developed by Northrop Grumman in the United States under the auspices of the Defense Advanced Research Projects Agency, illustrates the potential of next-generation UUVs to conduct prolonged deep-sea missions beyond the reach of conventional ship-based weapons, thereby substantially enhancing operational reach and mission persistence Harper (2024, May 1).

Within mine countermeasure (MCM) operations, UWS substantially enhance crew safety, accelerating the clearance of maritime passages. Specialized UUVs autonomously identify naval mines and deploy underwater charges to neutralize them effectively. A notable example is the Russian Surrogat UUV platform, which replicates the acoustic signature of a submarine to deceive adversary sensors and map minefields with high precision (Drone Wars UK, 2024). This approach enables more efficient use of operational resources while minimizing environmental harm by concentrating mine-clearance efforts on targeted zones.

The integration of UWS into the Distributed Maritime Operations (DMO) framework brings a profound shift in approaches to sea control and situational awareness. Autonomous platforms continuously collect operational data and transmit it to secure command nodes, where the information is processed by a combination of human analysts and AI-enabled decision-support systems to generate actionable targeting intelligence (Eckstein, 2024). This interconnected architecture enhances the operational resilience of naval task forces: even if individual UWS units are compromised, the system maintains functionality by dynamically reallocating tasks among remaining components. Such distributed robustness is especially valuable for states requiring rapid, adaptive responses in littoral zones.

Existing international norms do not provide a clear answer on how to ensure meaningful human control in complex operational environments, highlighting the need for an interpretative approach to legal frameworks. Simultaneously, the deployment of UWS encounters distinct technical and regulatory obstacles. Ensuring secure communication remains a critical concern, as adversaries may employ GNSS jamming or cyberattacks to sever control links Harper (2024, October 28). Furthermore, the physics of underwater communication creates inevitable latency, forcing UUVs to operate with high degrees of autonomy where real-time human intervention is impossible. This reality clashes with existing legal frameworks, as the question of accountability for outcomes generated by AI-enabled systems – based on human-defined models but executed independently – remains unresolved. This is especially critical in scenarios involving potential errors in target identification, underscoring the urgent need for a robust, context-sensitive interpretation of existing legal frameworks to maintain the standard of meaningful human control, even in communication-denied environments (Brock & Stone, 2024).

Despite these limitations, ongoing research continues to reveal tangible technological advancements. Companies such as Anduril and Shield AI, alongside initiatives by the U.S. Defense Innovation Unit, are developing next-generation platforms designed for extended endurance and rapid payload reconfiguration – from anti-submarine torpedoes to anti-ship strike systems (Katz, 2025a, 2025b). However, integrating such modular AI-enabled payloads requires rigorous legal assessments in accordance with Article 36 of Additional Protocol I to the Geneva Conventions. Compliance with IHL principles, particularly distinction and proportionality, dictates that the growing role of uncrewed maritime platforms should be understood not as a fully established operational standard, but as a developing capability whose lawful use depends on robust human oversight, transparent testing, and “lawful-by-design” architecture (Katz, 2025a).

Strategic Implications of AI Integration

Over the past five years, the integration of autonomous weapon systems has become a central topic in military, political, and academic discussions. These systems, which combine machine learning algorithms with high-speed sensor technologies, fundamentally transform the conventional decision-making chain, where the use of force was historically under the direct control of a human operator (Belfer Center, 2025). This shift generates two distinct yet interconnected strategic risks: the erosion of accountability frameworks and the potential for unintended escalation driven by “decision-time compression.”

The Accountability and Responsibility Gap

The deployment of AI-UxS blurs the boundaries between commanders, operators, and technical staff, creating uncertainty about who is responsible for specific outcomes. This transformation affects the entire lifecycle of the system, from design and coding to operational deployment. Under the jus in bello framework, legal responsibility for the use of force requires an identifiable individual or institutional actor, reflecting the principle of legal attribution, which ensures that a specific person, commander, or organization can be held accountable for actions carried out through autonomous or human-assisted systems, even if the system executes decisions independently or via algorithmic processes (Mazarr et al., 2020). A common counterargument holds that responsibility remains clear because the decision to deploy such a system itself establishes command responsibility for its foreseeable effects. While this position is legally persuasive in principle, it does not fully resolve practical attribution challenges, since highly autonomous systems may produce outcomes that are neither predictable nor directly controllable at the time of execution. When automated decision-support functions shape targeting decisions, assigning responsibility becomes legally and evidentially complex rather than conceptually absent. Failures stemming from algorithmic misjudgment, sensor malfunction, or cyber intrusions may result in incidental harm to civilians or civilian infrastructure. In such cases, distinguishing between strict legal liability and broader accountability (operational, organizational, and technical) is essential, particularly when the AI system’s underlying code is proprietary and inaccessible to independent review (Lewis, 2023).

The core difficulty lies not in the absence of applicable legal frameworks, but in translating legal obligations into technical requirements. The issue here pertains to the lack of any interdisciplinary vocabulary shared by programmers, manufacturers, and military commanders. For instance, if an AI-enabled decision-support system contributes to failures to comply with fundamental IHL obligations, such as distinction, proportionality, or the duty to take feasible precautions in attack, the problem is not a question of some legal vacuum: it relates to the erosion of human deliberative processes within the targeting cycle. Recent scholarship demonstrates that AI-DSS do not merely support human judgment; they reshape cognitive decision-making environments by accelerating operational tempo, structuring information flows, and promoting quantitative logics that risk reducing inherently qualitative legal assessments, particularly proportionality evaluations, to data-driven outputs (Dorsey, 2026; Dorsey & Bo, 2026). In such circumstances, the central legal concern shifts from the existence of command responsibility, which remains intact under IHL, to questions of accountability and liability allocation, especially where opaque algorithmic processes complicate retrospective scrutiny of how precautionary measures were evaluated or proportionality judgements were formed (Dorsey, 2026).

These discrepancies indicate that rather than replacing existing legal frameworks, regulatory efforts should prioritize interpretive refinement of IHL norms, more explicit doctrinal guidance on human–machine interaction within targeting processes, and complementary oversight mechanisms such as testing, evaluation, verification, and validation procedures integrated throughout the AI system lifecycle (Arms Control Association, 2024, November 12).

Decision-Time Compression and Escalation Risks

The second significant strategic implication concerns the speed of engagement. Data transmission and processing speeds in AI-UxS significantly exceed human capabilities, leading to a phenomenon known as decision-time compression. If a system identifies an object it classifies as a potential threat, it may initiate an engagement within milliseconds, long before a human operator has the opportunity to assess the situation or intervene effectively (Mazarr et al., 2020).

This extremely fast capability introduces the risk of unconscious escalation. When multiple autonomous systems are interconnected within a shared network or facing adversary autonomous systems, ultra-fast reactions can produce a cascading effect: an engagement by one unit may be interpreted by another as an act of aggression, prompting successive automated countermeasures (Rivera et al., 2024). This feedback loop effectively removes meaningful human control from the process, creating conditions where a localized tactical event can rapidly spiral into a broader strategic conflict without deliberate intent or political oversight. Without timely coordination of national and international efforts to manage these dynamics, there is a growing danger of strategic instability and the onset of a self-perpetuating AI arms race (Simmons-Edler et al., 2024).

Recommendations for State Policy

In the realm of national policy, creating favorable conditions for small, innovative enterprises through mechanisms of accelerated contracting can significantly reduce the time required to transition uncrewed systems from conceptual development to operational deployment. Drawing on the Defense Innovation Unit’s example, it is advisable to adopt simplified procurement procedures to facilitate rapid funding for prototype development and proof-of-concept phases. Such a framework prioritizes independent field-testing outcomes in decision-making for subsequent development stages. It reduces the development cycle from the initial idea to the deployment of a combat-ready model to six to nine months, compared to the conventional timeframe of eighteen to twenty-four months (U.S. Department of Defense, n.d.).

Modular budgeting, combined with phased milestone reviews, enables the formation of a flexible financial portfolio in which each subsequent funding tranche is allocated only upon the successful completion of the preceding stage (U.S. Department of Defense, n.d.; Center for Strategic and International Studies, 2021). This approach not only accelerates system development but also enables prompt adaptation of technical specifications to evolving battlefield conditions. In parallel, establishing a dedicated rapid-response reserve fund for financing emergent technological initiatives during periods of acute security challenges would minimize delays in deploying innovative solutions during crisis scenarios (Khalymon & Tyshchuk, 2026; Tyshchuk, 2024).

However, the development of public–private partnerships in the defense technology sector requires a carefully adapted model rather than a direct replication of the rapid, market-driven innovation cycles, minimal regulatory oversight, and risk-tolerant investment culture typical of Silicon Valley, which may conflict with military, ethical, and legal obligations. Given that the military applications of artificial intelligence directly affect civilian populations and protected infrastructure, such partnerships must strike a balance between innovation and stringent legal, ethical, and safety standards. A structured, multi-stakeholder framework should be established, involving universities, venture capital funds, defense contractors, legal professionals, military lawyers responsible for weapons reviews, and specialists in “responsible by design” methodologies. Their coordinated efforts are essential to ensure that technological development complies with the principles of international humanitarian law and established review mechanisms. Innovation clusters should therefore prioritize interdisciplinary oversight and controlled knowledge exchange over speed of implementation alone. Combining private investment with state support will facilitate the responsible advancement of dual-use technologies and help avoid bottlenecks during the transition from laboratory research to operational deployment (Center for Strategic and International Studies, 2021; Boussel, 2024, June 12).

National legislation must incorporate regulatory and legal instruments to govern the development and deployment of AWS, building on the foundational obligations set forth by IHL. These frameworks should be supplemented by ethical guidelines and transparency initiatives proposed by institutions such as the Centre for International Governance Innovation (CIGI), the Yale Initiative on the Regulation of Artificial Intelligence (YIRA), and the Campaign to Stop Killer Robots (CSKR). A key component of such regulation is the explicit requirement to maintain meaningful human control over any decision involving lethal force. This entails ensuring that a human operator retains the capacity to monitor the system in real time and to intervene or abort a mission if necessary (International Committee of the Red Cross, 2018; International Committee of the Red Cross, 2021, May 12; Csernatoni, 2021).

Comprehensive regulation must also encompass algorithmic transparency, the assignment of individual or institutional responsibility at every phase of the AWS lifecycle – ranging from system design and development to operational use – and the mandatory implementation of independent audits assessing software safety. These standards must be reflected in internal operational protocols and procurement standards for both military units and commercial suppliers (National Security Commission on Artificial Intelligence, 2021; Arms Control Association, 2024, November 12).

Guaranteeing human-in-the-loop (or “human-on-the-loop”) control involves ensuring that a human operator either authorizes each engagement or supervises the system’s behavior with the ability to intervene at any moment. These operational models serve as practical mechanisms for implementing the principle of both the legal standard of meaningful human control and the broader political formulation of context-appropriate human judgment and control (CAHJ&C) as articulated in the drafting text of the Group of Governmental Experts on Lethal Autonomous Weapons Systems; MHC provides greater doctrinal precision for assessing responsibility, while CAHJ&C offers political flexibility for state practices. In this context, the human retains ultimate authority over decisions involving the use of lethal force and must be able to immediately halt an AWS’s actions in the event of operational errors or target misidentification (U.S. Department of Defense, n.d.; Scharre, 2018). National legislation must articulate detailed requirements for algorithmic auditing, mandating that system decisions be interpretable by independent experts and subject to systematic annual review to verify compliance with IHL norms and to ensure the reliability and safety of autonomous systems (International Committee of the Red Cross, 2018; U.S. Department of Defense, n.d.).

The regulatory framework should be both modular and adaptable, combining binding legal norms rooted in international humanitarian law and domestic legislation with supplementary guidance documents and technical standards that can be regularly updated as new technologies emerge. To ensure regulatory responsiveness, it is advisable to establish “technology councils” within specialized legislative or governmental bodies. These councils would monitor global developments in AWS and promptly recommend updates to relevant rules and procedures (Center for Strategic and International Studies, 2021). Additionally, implementing regulatory sandboxes would facilitate controlled testing of new systems within designated test ranges. These environments would enable the assessment of operational safety, ethical implications, and technical performance, while promoting the structured inclusion of public stakeholders and domain experts in evaluating potential risks (OECD, 2023).

Effective governance of AWS requires robust interagency coordination. A specialized commission, including representatives from key ministries and research institutions, should be tasked with developing harmonized risk assessment standards for AWS and formulating guidelines for collaboration among military authorities, civilian institutions, and academic stakeholders (U.S. Government Accountability Office, 2020; NATO Standardization Office, 2022). Sustained information exchange with private AI developers and universities will enhance early detection of cybersecurity vulnerabilities and facilitate the design of preventive measures before the widespread deployment of AWS.

At the international level, active engagement in the Convention on Certain Conventional Weapons and its Group of Governmental Experts on Lethal Autonomous Weapons Systems should be intensified. Despite the participation of a limited number of states, this remains the principal United Nations forum for discussing regulatory norms on autonomous weapons. Ongoing collaborative efforts by state-appointed experts to develop shared understandings on regulatory approaches to autonomous weapon systems, as reflected in the working paper on draft regulatory measures (United Nations Office for Disarmament Affairs, 2025a) and the Chair’s summary of the 2025 session (United Nations Office for Disarmament Affairs, 2025b), are instrumental in advancing shared understandings of safeguards, human control, and legal review obligations. Such efforts contribute to harmonizing approaches to meaningful human control, reaffirm the requirement for legal review under Article 36 of the Additional Protocol I to the Geneva Conventions, and promote the exchange of information on national oversight mechanisms (United Nations Office for Disarmament Affairs, 2023; Spazian et al., 2021). Once adopted by consensus, these principles should be incorporated into national security and defense strategies, with appropriate adaptation to domestic legislative frameworks. In parallel, AI-enabled decision-support systems (AI-DSS), which shape targeting decisions without executing force directly, warrant regulatory oversight either within the existing CCW structure or, as some propose, through broader UN General Assembly processes to ensure inclusivity and wider state participation.

Collaboration with the United Nations Institute for Disarmament Research (UNIDIR) and other relevant UN bodies, particularly the Office for Disarmament Affairs (UNODA), should encompass active participation in the preparation of signpost papers and thematic studies on AWS. National representatives ought to be directly involved in drafting these documents, ensuring that their findings and recommendations are integrated into the process of updating domestic regulatory frameworks. In addition, supporting the United Nations initiative to introduce temporary moratoria on AWS testing in active conflict zones is a prudent measure aimed at reducing the risk of uncontrolled escalation and enhancing international stability (Spazian et al., 2021; United Nations Office for Disarmament Affairs, 2023).

Equally important is the formation of bilateral and multilateral working groups with partners from NATO, the European Commission, and AUKUS to align technical standards such as system interoperability, communication protocols, and cybersecurity measures for AWS, as well as to standardize safety assessment procedures. The integration of AWS-related provisions into NATO’s Framework for Emerging and Disruptive Technologies, as well as into mechanisms like the European Defence Fund (EDF) and Permanent Structured Cooperation (PESCO), will enhance interoperability among allied armed forces and support timely information sharing on operational incidents (U.S. Government Accountability Office, 2020; NATO Standardization Office, 2022).

Finally, to evaluate the practical risks of escalation cascades and the ambiguities surrounding accountability, it is recommended that multilateral research initiatives be initiated under the leadership of institutions with recognized expertise in automated escalation dynamics and cybersecurity, such as the RAND Corporation and the CyberPeace Institute. These organizations have conducted in-depth analyses of compressed decision-making timelines and near-miss incidents involving autonomous systems (Mazarr et al., 2020; Laird, B., 2020; Pangotra, 2024). Simultaneously, training programs for military legal professionals and cybersecurity experts should be developed in collaboration with leading academic institutions to cultivate specialists capable of assessing the legal, ethical, and technical dimensions of AWS deployment.