human-in-the-loop

Human-in-the-loop (HITL) is a design principle for autonomous and semi-autonomous systems—particularly weapons, surveillance, and decision-support tools—that requires a human operator to actively authorise critical actions before they are executed. In the security domain, the term is most often invoked in debates over lethal autonomous weapons systems (LAWS), where it distinguishes systems that require human approval for each engagement from "human-on-the-loop" systems (human supervises but the machine can act) and "human-out-of-the-loop" systems (fully autonomous).

The distinction has become central to discussions at the UN Group of Governmental Experts (GGE) on Lethal Autonomous Weapons Systems, convened under the Convention on Certain Conventional Weapons (CCW) since 2017. States including Austria, Brazil, and members of the Non-Aligned Movement have pushed for a legally binding instrument requiring "meaningful human control" over the use of force, while major military powers—the United States, Russia, the United Kingdom, and others—have generally preferred non-binding guidelines.

Key policy documents reference the concept directly. The U.S. Department of Defense Directive 3000.09, originally issued in 2012 and updated in January 2023, requires that autonomous and semi-autonomous weapons allow commanders and operators "to exercise appropriate levels of human judgment over the use of force." NATO's 2021 Artificial Intelligence Strategy and the REAIM (Responsible AI in the Military Domain) summits launched in The Hague in 2023 similarly emphasise human accountability.

Critics—including the Campaign to Stop Killer Robots coalition and the International Committee of the Red Cross (ICRC)—argue that nominal HITL arrangements can become superficial when operators face split-second decisions or rely on opaque algorithmic recommendations, a problem sometimes called automation bias. Proponents counter that HITL preserves international humanitarian law principles of distinction, proportionality, and accountability by ensuring a legally responsible human agent authorises each strike.

Beyond weapons, the term also appears in content moderation, border-control biometrics, and intelligence analysis, where it signals that algorithmic outputs are reviewed rather than auto-executed.