Military applications of AI are among the most consequential — and least visible — uses of the technology. Defense organizations invest heavily in AI for intelligence, logistics, cyber operations and weapons systems, while diplomats, scientists and civil society groups debate how far machines should be allowed to make decisions about the use of force.

Intelligence and analysis

Militaries and intelligence agencies collect far more imagery, video, signals and text than human analysts can review. AI helps by:

  • detecting and classifying objects such as vehicles, ships and aircraft in satellite and drone imagery;
  • transcribing and translating intercepted communications;
  • summarizing open-source information from news and social media;
  • fusing data from different sensors into a common picture.

The US Department of Defense’s Project Maven, launched in 2017, aimed to apply machine learning to drone video analysis. Google’s involvement led to employee protests in 2018, and the company declined to renew the contract — an early flashpoint in debates about technology companies working on military AI. Since then, many technology companies have taken on defense work openly, and some AI labs have revised policies to allow certain national-security uses.

Logistics and maintenance

Less dramatic but highly valuable: predicting equipment maintenance, optimizing supply chains, planning transport and managing personnel. These uses mirror civilian applications and are often the first to be adopted.

Cyber operations

AI assists in detecting intrusions, analyzing malware and finding vulnerabilities — capabilities useful both for defending networks and potentially for attacking them.

Drones and autonomy

The war in Ukraine has made drones central to modern warfare, with both sides using large numbers of inexpensive aerial and naval drones. Electronic jamming that breaks the link between drone and operator has driven interest in autonomous navigation and terminal guidance, where onboard software completes the final phase of a mission without a live connection. This brings the abstract debate about autonomous weapons into real battlefields.

The autonomous weapons debate

Lethal autonomous weapon systems (LAWS) are generally described as systems that can select and engage targets without further human intervention once activated. Positions differ:

  • Campaigners and many states call for a treaty prohibiting systems that cannot be meaningfully controlled by humans and regulating others, arguing that delegating life-and-death decisions to machines is unacceptable and risks violations of international humanitarian law.
  • Some major military powers prefer non-binding principles, arguing that existing law of armed conflict applies and that autonomy can improve precision.

Discussions have taken place for years under the UN Convention on Certain Conventional Weapons, and the UN General Assembly has adopted resolutions expressing concern and calling for further work.

National policies

The US Department of Defense’s Directive 3000.09, updated in January 2023, sets requirements for the design, testing and senior review of autonomous and semi-autonomous weapon systems, requiring that they allow commanders and operators to exercise “appropriate levels of human judgment over the use of force”. Other countries have published their own principles, and a US-led political declaration on responsible military use of AI and autonomy has been endorsed by dozens of states.

Key concerns

  • Accountability: who is responsible when an autonomous system causes unlawful harm?
  • Reliability: AI can fail unpredictably in situations different from training data, and adversaries actively try to deceive it.
  • Escalation: faster machine decision cycles could accelerate conflicts beyond human control.
  • Proliferation: cheap autonomous drones could spread to non-state actors.
  • Nuclear command and control: there is broad agreement that humans must retain control over nuclear weapons decisions.

Meaningful human control

Much of the debate centers on the idea of meaningful human control: that people should understand a system well enough, and have enough time and information, to make informed decisions about the use of force. Critics argue that a human merely approving machine recommendations in seconds is not meaningful control.

What to watch next

  • Negotiations at the UN on a possible instrument on autonomous weapons.
  • Battlefield use of autonomous drones and terminal guidance.
  • AI companies’ policies on defense and national-security work.
  • Military AI testing and evaluation standards.

Key takeaways

  • Militaries use AI for intelligence, logistics, maintenance, cyber operations and drones.
  • Project Maven triggered a 2018 debate about technology companies in military AI.
  • Jamming in Ukraine has driven interest in drones that complete missions autonomously.
  • States disagree on whether a treaty should prohibit or regulate lethal autonomous weapons.
  • Accountability, reliability, escalation and proliferation are central concerns.

Frequently asked questions

Are fully autonomous weapons in use?

Some systems, such as certain air defense and loitering munitions, can operate with high autonomy in specific modes. The degree of human control varies and is often not publicly disclosed.

Is there a ban on killer robots?

There is no specific international treaty banning lethal autonomous weapons. Discussions continue at the UN, and existing international humanitarian law applies.

Do AI companies work with militaries?

Increasingly, yes. Several AI and technology companies provide models and services for defense and national-security uses under their usage policies.