The debate on autonomous weapons has entered an important phase. The UN General Assembly’s 2025 resolution on lethal autonomous weapons systems, the 2026 Geneva discussions under the Convention on Certain Conventional Weapons and the Responsible AI in the Military Domain process all point to the same concern: beyond asking whether artificial intelligence will shape warfare, the world is now examining whether law can shape it in time.
This concern is immediate because emerging and disruptive military technologies are already moving from laboratory concepts into doctrines, procurement cycles and battlefield practices. Artificial intelligence, autonomous systems, quantum technologies, cyber capabilities, space-based platforms, unmanned systems and data-driven surveillance are changing how militaries observe, decide and strike. Their significance lies not only in better weapons, but also in a deeper transformation of conflict, deterrence and accountability.
The most serious legal and ethical question is the delegation of life-and-death decisions to machines. International humanitarian law rests on human judgment. Can distinction, proportionality, precaution, necessity and accountability be delegated to machines? Lethal autonomous weapons systems require context, restraint and responsibility. Human beings may create such systems, but their use must remain governed by law. A machine may classify a target, process data and react at speed, but it cannot carry moral burden or legal responsibility. Where human judgment is displaced, the laws of war begin to lose their centre.
This is why the concept of meaningful human control of AI matters. It is neither a diplomatic slogan nor a technical checkbox. It must become a substantive legal standard. Any system that can identify, select and engage targets without human intervention raises direct questions of accountability. If a civilian is wrongly targeted, if an algorithm misreads behaviour because of coding bias or design flaws, if training data is corrupted, or if commanders cannot explain how a system reached its decision, responsibility becomes blurred. In war, blurred responsibility is a legal failure more than a technical inconvenience.
AI is dual-capable. It can be enabling as well as dangerous, and a blanket ban on its military use would be impractical. AI can support logistics, early warning, defensive systems, demining, maritime safety, medical evacuation and disaster response. The real task is to separate permissible assistance from impermissible autonomy. Fully autonomous weapons that lack meaningful human control and cannot comply with international humanitarian law should be prohibited. Other autonomous and AI-enabled systems should be regulated through strict weapons reviews, audit trails, explainability requirements, human supervision and clear chains of responsibility.
Such a framework must also guard against a second danger: technological denial. Since emerging technologies are dual-use, the same capabilities that assist military targeting may also support agriculture, climate adaptation, public health, disaster management, infrastructure protection and civilian innovation. Export controls and technology restrictions framed in the language of security can easily become instruments of strategic exclusion. For developing states, the governance of military AI must therefore balance restraint with equitable access. Regulation should prevent misuse rather than freeze technological hierarchies.
This is where the Global South has an important stake. Much of the current debate is shaped by technologically advanced states that already possess the capabilities under discussion. Their preference for voluntary norms, national reviews and responsible-use declarations may preserve flexibility for them, while leaving others outside both innovation and rule-making. A legitimate governance framework cannot be written by a small club of technological powers. It must allow developing states to co-author the rules that will govern future battlefields.
Pakistan’s position is relevant in this regard. It has consistently supported a legally binding framework prohibiting fully autonomous weapons lacking meaningful human control, while also emphasising peaceful uses, equitable access and non-discriminatory technology governance. This is a balanced approach. It rejects the replacement of human judgment in lethal decisions, yet preserves the right of states to pursue peaceful and responsible technological development. That balance is essential for any serious legal architecture.
The global process, however, remains fragmented. The CCW provides an important forum, but its consensus-based structure slows progress. UNGA resolutions demonstrate growing political concern, yet they do not create binding law on their own. REAIM has widened the conversation by bringing states, industry, academia, civil society and technical experts into military AI governance, but responsible-use principles still need legal weight. The result is an emerging architecture made of declarations, rolling texts, national policies and expert processes. It requires consolidation.
A practical way forward is layered governance. First, states should negotiate a prohibition protocol on fully autonomous weapons that operate beyond meaningful human control. The precedent exists: the 1995 protocol banning blinding laser weapons showed that law can act before a technology becomes normalised on the battlefield. Second, states may develop regulatory obligations for other AI-enabled military systems, including legal review, testing, transparency, explainability and accountability. Third, a new multilateral institution on emerging and disruptive military technologies could track state practice, support capacity-building, issue interpretive guidance and help bridge the gap between technical complexity and legal responsibility.
Regional confidence-building measures are equally important. South Asia cannot afford an unregulated race in military AI. The region already carries the burden of nuclear deterrence, compressed decision timelines, contested attribution and recurrent crises. AI-enabled ISR, unmanned systems, cyber operations, precision strike and space-based surveillance could interact dangerously with existing crisis dynamics. Even semi-autonomous systems may deepen misperception if used in surveillance-heavy or first-strike-enabling roles.
A South Asian code of conduct on military AI would be a modest but useful beginning. It could reaffirm human control over lethal decisions, discourage the deployment of fully autonomous systems, restrict AI-enabled targeting in high-tension zones and encourage expert-level dialogue on doctrine, data governance and crisis communication. Such measures would not resolve the region’s disputes, but they could reduce the risk that machines accelerate a crisis beyond what humans can manage.
International law has often followed war. It has been written after devastation, after atrocities and after the consequences became impossible to ignore. Emerging military technologies do not allow that luxury. Once autonomous systems become embedded in doctrines and arsenals, reversing them will be far more difficult than regulating them now.
The next war may not begin with a formal declaration. It may begin with a misread signal, an automated response, a drone swarm, a corrupted dataset or a targeting system that acts faster than human judgment can intervene. If accountability is not built into law, design and doctrine, the future battlefield may be governed less by conscience than by code.
Killer algorithms need rules. Those rules must preserve innovation, protect peaceful access, prohibit autonomous killing beyond human control and ensure that the law of war remains human before the next war begins.
The writer is a former brigadier with expertise in deterrence, arms control and disarmament affairs.
