Autonomous Weapons Explained: What Exists and What Is Legal

Weapons that can select and strike a target without further human input already exist in limited forms. No international treaty specifically bans them. Their use is governed by the laws of war and by national policies such as the US Defense Department's Directive 3000.09. It is also governed by a United Nations process that reaches a decision point in November 2026.

What is an autonomous weapon?

There is no single agreed definition. The International Committee of the Red Cross (ICRC) uses the term for any weapon that selects and applies force to targets without human intervention. A person activates the system, the ICRC says. Sensors and software then trigger a strike, it says, when what they detect matches a preset target profile, such as the shape of a military vehicle.

US policy sorts weapons by the role of the human operator, according to a Congressional Research Service primer updated in March 2026:

• Semi-autonomous (human in the loop): engages only targets a person has selected, as certain fire-and-forget guided missiles do.
• Human-supervised (human on the loop): can select targets itself, but an operator can monitor it and halt an engagement.
• Autonomous (human out of the loop): once activated, selects and engages targets with no further operator input.

Are autonomous weapons being used today?

In narrow roles, yes. A 2022 ICRC explainer describes mines as rudimentary autonomous weapons. As examples, it lists air defense systems that strike incoming missiles and some loitering munitions built to destroy radars or armored vehicles. It says these have generally been used against clear military targets, in areas with few civilians, and under close human supervision.

The war in Ukraine has gone further. IEEE Spectrum reported in March 2026 on one company, The Fourth Law. According to the report, the company says it has sent troops in eastern Ukraine thousands of low-cost modules that steer a drone through its final approach to a target. IEEE Spectrum also reported that Ukraine has found AI-capable chipsets in the wreckage of Russian-launched Shahed drones.

The report quoted Kate Bondar of the Center for Strategic and International Studies. She said current software cannot tell a soldier from a civilian. She also said she expects humans to stay in the loop for some years.

Senior US defense officials have not publicly confirmed whether the country is developing or has developed such weapons, the CRS primer says.

Are autonomous weapons legal?

No treaty prohibits autonomous weapons as a class. But international humanitarian law, the rules governing armed conflict, applies to them. The final report of a UN expert group was adopted on September 4, 2026. The report states that this law applies fully to the development and use of such systems. It says they may be used only in line with the principles of distinction, proportionality and precautions in attack. It also says human control and judgment are needed to comply.

The ICRC argues that is not enough. Since 2021 it has recommended new binding rules. These would prohibit unpredictable autonomous weapons and those designed to target people, and regulate the rest.

US policy does not ban the technology. Directive 3000.09 was last updated in January 2023. It requires that systems let commanders and operators exercise appropriate levels of human judgment over the use of force. CRS notes that this does not mean manual control. Covered systems need sign-off from three senior officials before development and again before fielding. The Deputy Secretary of Defense can waive that review for urgent military need. Congress must now be notified of any waiver.

Rules for the AI inside such systems are thinner. A CRS brief covers the Pentagon's dispute with AI company Anthropic and was updated on October 5, 2026. It says Congress has not legislated on the department's use of AI models. It also says the department is not publicly known to use any frontier AI model in autonomous weapons.

Where do treaty talks stand?

The UN General Assembly adopted a resolution on lethal autonomous weapons systems on December 1, 2025. The vote was 164 to 6, with 7 abstentions. The United States, Russia and Israel were among those voting against, and China abstained. The text called on parties to the Convention on Certain Conventional Weapons (CCW) to work toward completing a set of elements for an international instrument. It said this was with a view to future negotiations.

A CCW Group of Governmental Experts spent three years on that drafting. Its final report recommends that member states consider and endorse the elements at a review conference on November 16 to 20, 2026. It also recommends that they decide there on possible future work.

Human Rights Watch, which campaigns for a treaty, said states including Russia and the United States insisted on changes that weakened the text. The group counts 76 countries as supporting negotiations.

On August 25, 2026, UN Secretary-General António Guterres and ICRC President Mirjana Spoljaric jointly urged states to use the conference to move to negotiating a legally binding instrument.

What to watch

The CCW review conference is the next test. The convention works by consensus, so opposition from a few governments could prevent a negotiating mandate. In Washington, the Pentagon owes Congress annual reports through 2029 on the approval and deployment of lethal autonomous weapon systems.