Artificial intelligence is changing the way wars are fought, from analysing intelligence and identifying potential targets to guiding drones and supporting military planning. As these technologies become more capable, the United Nations is stepping up efforts to address the risks and ensure that human judgement remains central to decisions involving the use of force.
A new initiative led by the UN Institute for Disarmament Research (UNIDIR) is bringing governments, technology companies, researchers and security experts together to examine how AI is being used in military operations. The goal is to identify practical safeguards, encourage responsible industry practices and build a clearer understanding of the threats these systems could pose to international security.
The initiative comes at a time when military organisations are increasingly turning to AI to process battlefield information and support operational decisions. While these tools can help analyse large amounts of data quickly, they also raise difficult questions about reliability, accountability and how much authority should be given to automated systems.
UNIDIR Director Robin Geiss has warned that human judgement risks being pushed aside as military decision-making becomes increasingly automated. The concern extends to nuclear security, where a mistaken assessment or a rushed decision could have consequences far beyond the battlefield.
The issue is not simply whether AI can make military operations faster or more efficient. It is whether commanders can understand the systems they rely on, recognise their limitations and remain accountable when their recommendations influence decisions that could cost lives.
The initiative will also examine how armed groups outside government control might use increasingly accessible AI tools for operational planning, coordination and logistics.
For the United Nations, the challenge is to turn growing international concern into practical safeguards. Countries may agree that military AI carries serious risks, but they remain divided over how far restrictions should go and whether new rules should be legally binding.
WHY MILITARY AI IS BECOMING A GLOBAL SECURITY CONCERN
AI is already changing how armed forces gather information, assess threats and coordinate operations. Systems capable of analysing satellite images, surveillance footage and battlefield reports can help commanders identify developments that might otherwise take hours to assess.
The technology also has potential applications in drone navigation, reconnaissance, logistics and cyber defence. Used carefully, these tools may improve efficiency and help military personnel make better-informed decisions.
But speed does not always mean accuracy.
An AI system could flag a building, vehicle or individual as a potential threat based on incomplete or misleading information. If military personnel act on that assessment without adequate verification, the result could be a mistaken strike or civilian casualties.
The risks become more serious when operators have only seconds to review a recommendation. Under pressure, there may be little opportunity to question the system's conclusions or investigate whether its underlying assumptions are correct.
Another problem is that some AI systems are difficult to interpret. Their recommendations can be influenced by complex data patterns that are not immediately obvious to the people using them. This makes rigorous testing, clear operating limits and meaningful human supervision especially important.
These concerns extend beyond weapons themselves. AI-assisted intelligence assessments and threat warnings can influence military decisions long before a weapon is deployed.
When such systems are connected to nuclear warning or command structures, the consequences of an error could be particularly severe.
THE UNITED NATIONS PUSHES FOR STRONGER AI OVERSIGHT
Bringing governments, industry and researchers together
UNIDIR's initiative aims to create a space where governments, defence technology companies, researchers and civil society can examine the security implications of military AI and identify areas for cooperation.
Three priorities stand out.
Responsible industry practices: Technology companies can help reduce risks by testing systems thoroughly, documenting their limitations and designing safeguards that allow human operators to intervene when necessary.
Common technical language: Countries need a clearer understanding of terms such as autonomy, reliability, safety and human control. Without shared definitions, international discussions can become difficult to translate into workable policies.
Preventing misuse: The initiative will also consider how armed groups outside state military structures might use AI to support operational planning, coordination and logistics.
The involvement of private companies is particularly important. Many of the software tools, computing systems and AI capabilities used in modern defence are developed by commercial organisations.
Choices made during development can influence how a system performs under pressure, how its limitations are communicated and whether operators can override its recommendations when something goes wrong.
Voluntary industry guidance could help establish common expectations. However, its impact will depend on whether companies follow those standards and whether governments incorporate them into procurement, testing and military oversight.
Why the timing matters
International discussions about autonomous weapons and military AI have been underway for years. Yet the technology continues to develop, creating new questions about how existing laws and safeguards should be applied.
In August 2026, UN Secretary-General António Guterres and International Committee of the Red Cross President Mirjana Spoljaric renewed their call for international rules governing autonomous weapons. They urged states to prohibit unpredictable systems and weapons designed to target humans, while advancing negotiations towards a legally binding international instrument.
The latest UN initiative adds a practical dimension to that broader effort. Rather than focusing exclusively on weapons that can select and attack targets without direct human intervention, it considers AI's wider role in military decision-making.
That distinction matters because AI can shape the course of an operation well before a weapon is fired. Intelligence analysis, threat classification and automated recommendations can all influence how commanders interpret a situation and decide what to do next.
NUCLEAR SECURITY LEAVES LITTLE ROOM FOR ERROR
Few areas of military AI raise greater concerns than nuclear decision-making.
Nuclear-armed states must be able to assess warnings of a possible attack under intense pressure. AI could help process large volumes of information more quickly, but unreliable data or a misleading recommendation could also complicate decisions that already carry enormous consequences.
Imagine a system identifying a pattern that appears to indicate an imminent attack. If the information is incomplete or the pattern has another explanation, the warning could be wrong. If decision-makers place too much trust in the system, they might feel pressure to act before independent checks are complete.
The danger lies not only in a system making a mistake, but also in the possibility that people may have too little time to recognise it.
This does not mean AI independently controls nuclear arsenals. The concern is that greater reliance on automated analysis could narrow the time available for human assessment and increase the risk of miscalculation during a crisis.
Keeping human authority firmly in place is therefore essential when decisions involve nuclear weapons.
International cooperation could help establish expectations for testing, verification, crisis communication and the use of AI in sensitive military systems. Reaching agreement will be difficult, however, because governments have competing security interests and are often reluctant to reveal details about classified military capabilities.
AUTONOMOUS WEAPONS RAISE QUESTIONS ABOUT HUMAN CONTROL
Autonomous weapons remain at the centre of the international debate over military AI.
Depending on their design, these systems can carry out certain functions with limited human intervention, including identifying, selecting or engaging targets using sensor data and predefined operating parameters.
The central question is how much meaningful control a person retains over the decision to use lethal force.
International humanitarian law already requires parties to armed conflicts to distinguish between civilians and combatants, assess proportionality and take feasible precautions to protect civilians. Introducing AI into military operations does not remove these responsibilities.
However, determining accountability when an automated system causes harm can be complicated. Investigators may need to establish whether a failure resulted from faulty design, inadequate training data, unexpected operating conditions, an operator's decision or a lack of proper supervision.
Clear responsibility matters because the use of increasingly sophisticated technology must not create gaps in legal accountability.
The UN and the ICRC have called for clear prohibitions and restrictions on autonomous weapons. The seventh Review Conference of the Convention on Certain Conventional Weapons, scheduled for November 2026, is expected to provide an important opportunity for governments to consider further steps towards international rules.
Progress remains uncertain. Countries disagree over what should count as an autonomous weapon, which applications should be restricted and whether new requirements should be legally binding.
GLOBAL MILITARY COMPETITION COULD SLOW PROGRESS
Efforts to regulate military AI are unfolding alongside an intensifying race to develop advanced defence technologies.
The United States, China, Russia and other military powers have strong incentives to improve their intelligence capabilities, autonomous systems and operational speed. Governments may fear that strict restrictions could weaken their position if rivals continue developing similar technologies.
This creates a difficult balancing act. Countries may recognise the dangers of poorly controlled military AI while resisting measures they believe could limit their strategic advantage.
Even so, some risks create opportunities for cooperation. In September 2026, the United States proposed an AI-related notification mechanism with China intended to help the two countries communicate about incidents that could threaten national security.
Such a mechanism could help reduce misunderstandings between major powers, although it would not replace broader rules governing autonomous weapons or the use of AI in military operations.
For the UN, an immediate priority is to identify areas where countries can make progress despite their differences. Shared technical standards, improved communication during crises and stronger testing requirements could provide a starting point while negotiations over binding restrictions continue.
The larger challenge will be ensuring that safeguards remain effective as military technology evolves.
WHAT TO WATCH NEXT
The new UN initiative: Watch for its first recommendations on military AI, human oversight and responsible industry practices.
The November autonomous-weapons review conference: Any progress towards negotiations on a legally binding international instrument could shape future rules.
US–China security discussions: Developments in the proposed AI incident-notification mechanism could offer a practical model for reducing misunderstandings between military powers.
Industry commitments: The adoption of stronger safety testing, transparency measures and human-control safeguards will help show whether voluntary standards can make a meaningful difference.
Nuclear command and warning systems: Any new policies addressing AI-assisted threat assessment and human decision-making will deserve close attention.
AI use by armed groups: Evidence of AI being used for operational planning, cyber activities or logistics outside conventional state military structures could increase pressure for wider safeguards.
International legal standards: Watch for concrete proposals defining unacceptable systems, setting limits on autonomy and establishing minimum requirements for human control.
FACT CHECK METER
98% — STRONGLY VERIFIED
Assessment based on the UN, UNIDIR and international reporting cited in the original article. The percentage is an editorial assessment, not an independently calculated statistical measure.
Confirmed details
UNIDIR is pursuing an international initiative focused on military AI and international security.
UNIDIR Director Robin Geiss has raised concerns about human judgement being squeezed out of military decision-making.
The initiative addresses responsible industry practices, shared technical terminology and the potential misuse of AI by armed groups.
The UN Secretary-General and the ICRC president have renewed calls for international rules governing autonomous weapons.
Military AI was discussed at a UN Security Council meeting on 23 September 2026.
Important distinctions
The new initiative is a forum and policy-development effort, not a newly enacted global treaty.
Voluntary industry guidance is not the same as legally binding international restrictions.
AI-assisted military systems are not all autonomous weapons; the degree of human involvement varies.
Concerns about nuclear decision-making relate to the potential risks of AI integration, not confirmation that AI independently controls nuclear weapons.
AWX Verdict: The UN's initiative represents a significant step in the international debate over military AI. The real test will be whether governments and industry can turn broad concerns into practical safeguards, clear responsibilities and enforceable rules while competition in defence technology continues.
THE AWX BOTTOM LINE
Artificial intelligence is becoming an increasingly important part of modern warfare. It can help armed forces process intelligence, identify potential threats and coordinate operations, but it also introduces new risks when decisions are made too quickly or systems are trusted without sufficient scrutiny.
The United Nations is trying to bring governments, researchers and technology companies together before those risks become harder to manage. Its focus is on protecting human judgement, strengthening accountability and reducing the chance that an AI-related error could lead to civilian casualties or a wider military escalation.
Getting countries to agree on meaningful restrictions will not be easy. Governments must weigh the potential advantages of military AI against the consequences of losing control over decisions involving human life.
The November review conference on autonomous weapons could offer an important opportunity to move the debate forward. But progress will depend on whether countries are prepared to agree on specific limits rather than relying on general promises of responsible development.
The bottom line: Military AI is advancing rapidly, while international safeguards are still taking shape. The UN's challenge is to help close that gap and ensure that technology does not weaken civilian protection, increase the risk of conflict or blur human responsibility for the use of force.
ATTLAS NEWS | AWX WORLD NEWS
Real Stories. Lasting Impact.
Comments
No comments yet.