How Will Military AI and Hybrid Warfare Be Regulated?
Introduction: The Regulatory Gap
The rapid integration of artificial intelligence into military systems and the proliferation of hybrid warfare tactics have created a regulatory environment defined less by coherent law than by strategic ambiguity. States are developing autonomous weapons capable of selecting and engaging targets without direct human intervention, while simultaneously conducting cyber operations against critical infrastructure and deploying algorithmic influence campaigns that blur the boundary between peace and war. Existing international law—particularly International Humanitarian Law (IHL)—was not designed for weapons that decide, or for conflicts fought through code and cognition rather than kinetic force. The central question is not whether military AI and hybrid warfare can be regulated, but whether the political will exists to close the gap between legal principles and operational realities. Current trends suggest that regulation will proceed through fragmented, dual-track processes: binding restrictions on the most extreme autonomy, voluntary norms for cyber operations, and an expanding but contested role for IHL in the “grey zone.”
Regulating Lethal Autonomy: The Battle Over Meaningful Human Control
The most concrete regulatory debate concerns lethal autonomous weapon systems (LAWS). The prevailing consensus among most states, the International Committee of the Red Cross (ICRC), and the UN Secretary-General is that meaningful human control must be preserved over decisions to use lethal force. The question is how to translate this principle into binding rules. Switzerland and Greece have advocated a “two-tier approach”: systems that cannot comply with IHL—because they are inherently indiscriminate or uncontrollable—would be prohibited, while those capable of compliance would be subject to strict regulations, including limits on target types, duration, and geographic scope, with operators retaining the ability to abort missions.
However, negotiations within the UN Convention on Certain Conventional Weapons (CCW) have exposed deep divisions. In September 2026, reports emerged that US and Russian diplomats pushed to delete provisions requiring AI weapons to operate in “predictable and reliable” ways, to consider ethical factors, and to ensure human review of AI-designated targets before strikes. The closed-door nature of these modifications and the involvement of large legal delegations suggest that major powers are positioning themselves to preserve operational flexibility rather than accept binding constraints. This pattern mirrors earlier failures in the CCW process: the 2021-2025 Open-Ended Working Group concluded without carrying forward reaffirming language on the application of international law to cyber operations.
The legal difficulty extends to accountability. As Guatemala noted in UN deliberations, creating a lethal system without human control may violate the right to life and IHL by making criminal responsibility impossible to assign—an abstract algorithm cannot be prosecuted. Article 36 of Additional Protocol I requires states to review new weapons for IHL compliance, but the complexity and unpredictability of autonomous systems make such reviews inherently challenging. The likely trajectory is not a comprehensive treaty but a patchwork of national review mechanisms, export controls, and politically binding declarations, with the strongest prohibitions applying only to systems that lack any human override.
Cyber Operations and the Limits of IHL Application
For state-sponsored cyber warfare against critical infrastructure, the regulatory framework is more developed in principle than in practice. UN Groups of Governmental Experts have affirmed since 2013 that international law, including the UN Charter, applies in cyberspace. The ICRC has clarified that IHL governs cyber operations during armed conflict, prohibiting attacks on civilian objects such as hospitals, power grids, and water systems, and requiring that indiscriminate cyber tools not be used. The Tallinn Manual provides scholarly consensus that cyber operations amounting to “attacks” must comply with distinction, proportionality, and precaution.
Yet enforcement remains largely absent. The NotPetya attack of 2017, which caused billions in global damage, was publicly attributed to Russia but never examined in any venue capable of establishing legal responsibility. The SolarWinds compromise similarly highlighted how disagreements over the legal status of cyber espionage limit the practical application of law. The fundamental problem is institutional: the International Court of Justice can only hear cases with state consent, which is rarely given, while UN working groups lack mandates to adjudicate specific allegations.
A significant doctrinal debate concerns whether cyber operations that cause functional disruption—deleting data, disabling networks—constitute “attacks” under IHL. Some states, including China, argue that physical damage or casualties are required; others contend that disabling critical systems can have humanitarian consequences equivalent to kinetic strikes. This ambiguity creates space for operations that inflict severe harm while remaining below the legal threshold. The ICRC has proposed four rules for civilian hackers during armed conflict, drawing on existing IHL principles, but these are voluntary and lack enforcement mechanisms. States have obligations to prevent their territory from being used for internationally wrongful cyber acts—the due diligence principle—but its scope remains contested, particularly regarding what constitutes sufficient “knowledge” and “capacity.”
The most plausible regulatory development is not a cyber treaty but continued elaboration of voluntary norms, confidence-building measures, and the newly established UN Global Mechanism, which began operations in 2026. These mechanisms can improve transparency and communication but are unlikely to deter determined state actors.
Hybrid Warfare and the Cognitive Domain: The Least Regulated Frontier
The most acute regulatory gap concerns hybrid warfare, particularly AI-driven influence operations and cognitive warfare. Disinformation, deepfakes, and algorithmic manipulation are used to destabilize societies, interfere with elections, and undermine military decision-making. Yet IHL is heavily oriented toward physical violence and observable destruction, leaving non-kinetic psychological harm only partially addressed. The legal status of disinformation as a “means or method of warfare” remains unsettled, and the threshold at which cognitive operations become legally relevant is undefined.
The problem is compounded by attribution. AI-generated content can be produced and disseminated through automated systems, platforms, and users without a clear or singular author. This complicates the application of state responsibility, which requires establishing that conduct was carried out on the instructions of, or under the direction or control of, a state. The involvement of private technology companies in developing and controlling the infrastructure of cognitive warfare further fragments accountability.
Some scholars propose an effects-based interpretation of IHL: when cognitive operations are sufficiently connected to hostilities, they should be governed by existing protective rules, with precautionary controls built into design and deployment. Others argue for a broader reinterpretation that treats serious psychological harm and decision-making interference as legally cognizable damage. The ICRC has begun to address these questions, but states remain far from consensus.
Conclusion: Fragmentation as the Status Quo
The regulation of military AI and hybrid warfare will not follow a single, coherent trajectory. Instead, it will proceed along parallel tracks with different speeds and bindingness. For lethal autonomy, the likely outcome is a weak CCW instrument or politically binding declaration, supplemented by national review processes and export controls—with major powers resisting strict limits. For cyber operations, the existing framework of voluntary norms and confidence-building measures will expand incrementally, but enforcement will remain elusive absent a venue with jurisdiction and state consent. For cognitive warfare, the regulatory gap will persist, with IHL stretched—often uncomfortably—to cover domains it was never designed for.
The deeper obstacle is not legal but political. As the US-Russia maneuvers in the LAWS negotiations demonstrate, states with significant military AI capabilities have little incentive to accept binding constraints. The “grey zone” is not a legal vacuum but a space deliberately constructed to avoid legal accountability. Regulation will advance only when the costs of ambiguity—escalation, miscalculation, civilian harm—become sufficiently visible to outweigh the strategic advantages of deniability. Until then, international law will lag behind the technologies it seeks to govern, and states will continue to exploit the gap between principle and enforcement.