T
22 November 2026 · 3 views

Red Cross Warns Time to Regulate Autonomous Weapons Is Short

Red Cross Warns Time to Regulate Autonomous Weapons Is Short

The International Committee of the Red Cross (ICRC) warns that states must act urgently to preserve meaningful human control over decisions involving the use of force. As artificial intelligence, sensors and uncrewed platforms advance, military systems may increasingly detect, select and attack targets with limited human intervention.

The ICRC says future warfare must not be governed by systems that operate beyond meaningful human judgment and responsibility. Source 1

What Are Autonomous Weapons?

Autonomous weapons are military systems capable of performing critical functions without continuous human direction. Depending on their design, these functions may include detecting objects, identifying potential targets, selecting targets and initiating or applying force.

Autonomy exists on a spectrum. A remotely controlled drone may transmit images to a human operator who decides whether and where to strike. An autonomous system may instead search an area, identify objects according to programmed criteria and engage targets after activation.

The central question is whether people retain sufficient understanding, judgment and authority over the use of force. Automation does not automatically make a weapon unlawful or uncontrollable, but human involvement becomes inadequate when a system acts too quickly, behaves unpredictably or processes information that an operator cannot realistically evaluate.

How Artificial Intelligence Is Changing Military Systems

Artificial intelligence, machine learning and advanced sensors are expanding the capabilities of:

  • Uncrewed aerial vehicles and drones.
  • Missile defence systems.
  • Uncrewed ground and maritime vehicles.
  • Surveillance and targeting platforms.
  • Swarm systems involving multiple coordinated machines.
  • Software that analyzes large volumes of battlefield information.

These technologies may help personnel identify threats or operate in dangerous environments. The principal concern is the delegation of critical decisions to systems operating in uncertain, chaotic or difficult-to-interpret conditions.

A system tested in controlled conditions may encounter unfamiliar terrain, unusual objects, civilian movement or unexpected changes during an armed conflict. If it can select and attack targets without adequate human review, errors may occur at high speed and across a wide area.

Why Regulation Is Urgent

Technology Is Developing Faster Than International Rules

The ICRC warns that states may lose the opportunity to establish effective rules before autonomous weapons become more widespread. Source 3

Military technology can progress rapidly, while international negotiations often take years. Once systems are deployed, governments may become dependent on them, making restrictions harder to accept. Early standards can shape research, design, testing, procurement and deployment before practices become established.

Meaningful Human Control Must Remain Central

Meaningful human control requires more than placing a person somewhere in an approval chain. Decision-makers need enough information, time and authority to make an informed judgment about the use of force.

Effective control may require:

  • A clear understanding of the weapon’s capabilities and limitations.
  • Reliable information about the target and surrounding environment.
  • Enough time to assess the situation.
  • Defined geographic, temporal and operational limits.
  • The ability to intervene, cancel or stop an attack.
  • Clear responsibility for authorizing force.
  • Monitoring during the weapon’s operation.

The ICRC argues that states must preserve human control as autonomous weapons develop. Source 3

States Must Retain Responsibility

Responsibility for military force cannot be transferred to an algorithm, software developer, contractor or autonomous platform. Machines cannot bear legal or moral responsibility for their actions.

The ICRC president has urged states to ensure that the future of warfare is not delegated to machines. Source 9

Responsibility must extend across the weapon’s life cycle, including research, design, testing, procurement, deployment, operation and investigation. Commanders and officials must know who authorizes a mission, who supervises it and who can suspend it when circumstances change.

Legal Risks

International Humanitarian Law

International humanitarian law requires parties to distinguish between civilians and combatants, and between civilian objects and military objectives. Weapons must be used in ways that comply with these and other obligations.

Autonomous systems may struggle with ambiguous situations, including recognizing civilians, people who are surrendering or wounded, and objects with both civilian and military uses. Performance may also change because of poor visibility, damaged infrastructure, unexpected behavior or incomplete information.

Before deployment, states must assess whether a weapon can operate consistently with international humanitarian law. That assessment must consider realistic environments, likely failures and the consequences of uncertainty, rather than relying only on controlled demonstrations.

Proportionality and Precautions

The principle of proportionality prohibits attacks expected to cause civilian harm excessive in relation to the anticipated military advantage. Parties must also take feasible precautions to reduce civilian harm.

Conditions can change between target identification and impact. Civilians may enter an area, a target may move, intelligence may become unreliable or the expected military advantage may diminish. An autonomous system may therefore need to reassess targets, identify unreliable information, abort attacks and adjust its response.

The more complex the environment and the faster the system acts, the harder it becomes for humans to evaluate or correct its decisions.

Accountability Gaps

An unlawful attack involving an autonomous weapon could involve political leaders, commanders, procurement officials, designers, operators and manufacturers. Without clear rules, responsibility may become fragmented.

States need documented decisions, clear chains of command, reliable records and procedures for investigating incidents. Assigning responsibility after deployment cannot replace effective controls before deployment. Human institutions must remain responsible for deciding whether, when and how a system is used.

Ethical Concerns

Delegating decisions to kill may reduce direct human responsibility and weaken the moral judgment associated with the use of force. Humans can interpret context, recognize vulnerability and respond to surrender or changing circumstances. A machine processes inputs according to its design and does not possess human moral agency.

AI systems also depend on data, models, sensors and assumptions that may contain errors or bias. Smoke, camouflage, damaged infrastructure, civilian movement and incomplete information can undermine performance. Testing can reduce risk but cannot guarantee correct behavior in every operational situation.

Autonomous weapons may also lower the perceived political cost of using force by reducing risks to a state’s own personnel. Faster operations, larger-scale attacks and more difficult-to-control escalation create strategic and political risks beyond technical safety.

International Negotiations

United Nations discussions address lethal autonomous weapons and systems that can select and attack targets with limited human control. Negotiators are considering whether some systems should be prohibited and others strictly regulated. Source 5

Key questions include:

  • Which systems should be prohibited?
  • Which systems could operate under strict controls?
  • What level of human supervision is required?
  • How should states test and review weapons?
  • How can compliance be verified?
  • Which definitions should apply to autonomous functions?

UN Secretary-General António Guterres and the ICRC have renewed calls for international rules. Source 7

Agreement is difficult because states have different security priorities and may disagree over definitions. Verification is also complex: software, sensors and mission parameters can change system behavior. Terminology may delay substantive progress while the technology continues to develop.

Possible International Rules

One approach would prohibit weapons that select and attack human targets without meaningful human control. Rules could also prohibit systems that operate unpredictably in populated areas or cannot be reliably constrained by location, duration, target type or mission purpose.

Other autonomous systems could face strict limits, including:

  • Narrow target categories.
  • Restricted geographic areas.
  • Short operating periods.
  • Defined mission parameters.
  • Human authorization before lethal force is used.
  • Reliable cancellation and override mechanisms.
  • Limits on operation near civilians.
  • Traceable records and post-operation review.

States could also require realistic testing, documentation of capabilities and limitations, assessment of human-machine interaction, verification of override functions, clear command responsibility, incident reporting and reassessment after software, sensor or mission changes.

What States Should Do Now

States should establish clear national policies defining acceptable and prohibited autonomous functions. Policies should identify who authorizes deployment, supervises operations, suspends missions, investigates incidents and reviews system updates.

They should preserve human decision-making by ensuring that operators have sufficient information, time, training and practical authority to intervene. Rules should cover planning, deployment, supervision, review and investigation.

States should also participate in United Nations negotiations and support shared terminology and minimum standards for testing, safety controls, incident reporting and weapons reviews. Binding rules would provide stronger protection than voluntary principles alone.

Why Delay Makes Regulation Harder

Once autonomous weapons are integrated into command structures, logistics and military doctrine, restrictions may face stronger resistance. Governments may be less willing to abandon systems already in service or redesign procedures built around them.

Repeated use could also normalize practices before international law clearly addresses them. Early deployment may influence procurement decisions, state behavior and public acceptance. Regulation is most effective before systems are widely deployed, when it can still shape design choices, testing standards and operating procedures.

Conclusion

Autonomous weapons regulation is both a technological and political decision. The ICRC warns that states must act before machines become more deeply involved in decisions about the use of force.

The risks include civilian harm, legal uncertainty, accountability gaps, unpredictable behavior and reduced human responsibility. Existing international humanitarian law remains relevant, but states are debating whether additional rules are needed for weapons that can select and attack targets autonomously.

Possible measures include prohibiting systems that cannot be meaningfully controlled, restricting other autonomous weapons by target and environment, requiring human authorization for lethal force, and strengthening testing and oversight.

The future of warfare should not be determined solely by what technology can do. International rules must establish where machines may assist military operations and where human beings must retain control.

Frequently Asked Questions

What are autonomous weapons?

Autonomous weapons are military systems capable of detecting, selecting or attacking targets with limited human intervention. Their autonomy varies according to the system, mission and operating environment.

Why is the Red Cross warning about autonomous weapons?

The ICRC warns that states may soon lose the opportunity to establish effective rules before autonomous weapons become more advanced and widely deployed. It calls for urgent action to preserve meaningful human control.

What does meaningful human control mean?

It requires people to understand a weapon’s capabilities and limits, make informed decisions about its use, supervise its operation and retain the ability to intervene or stop an attack.

Are autonomous weapons already regulated by international law?

Existing international humanitarian law applies to the use of weapons and includes requirements concerning distinction, proportionality and precautions in attack. States are debating whether additional rules are needed.

What are “killer robots”?

“Killer robots” is a common term for lethal autonomous weapons that can identify and attack targets without meaningful human control. International discussions may instead use terms such as autonomous weapons systems or lethal autonomous weapons systems.

What rules are states considering?

Potential measures include prohibiting systems that cannot be meaningfully controlled, restricting autonomous weapons to specific targets or environments, requiring human authorization for lethal force, and strengthening testing, oversight and accountability.

3 views