The Royal Navy launched one-way attack drones from the aircraft carrier HMS Queen Elizabeth for the first time, BAE Systems and the Navy announced on 5 November. Four British-made Nyan One-Way Effector aircraft, built by Callen-Lenz in Wiltshire, flew from the carrier during Exercise Neptune Reach B in the North Atlantic in October, and the Navy says all four reached their designated strike points. The trial is the clearest sign yet of Britain's plan to turn carriers designed around the crewed F-35B into "hybrid air wings" that mix expensive fighters with cheap, expendable drones.
The logic is cost. An F-35B is scarce, needs a trained pilot and demands heavy maintenance; a one-way effector is meant to reach a target once and not come back. The Navy is testing whether pairing the two lets commanders put more weapons in the air without buying more jets. Separate trials aboard the same carrier flew the reusable Tekever AR5, a seven-metre-wingspan surveillance and logistics drone, from Portugal to the ship and back, filling the reconnaissance and resupply roles a strike drone cannot.
Britain is not alone, and the trend runs toward weapons that decide more for themselves. In July, the US firm Anduril's YFQ-44A "Fury" collaborative combat aircraft fired an air-to-air missile at a simulated target over California, with a human approving the shot and the AI carrying it out; in October an F-22 pilot directed the drone through simulated combat tasks. The Pentagon's roughly $1.1 billion Drone Dominance program, which implements a memo by War Secretary Pete Hegseth, is buying low-cost drones through competitive "Gauntlet" fly-offs and aims for more than 200,000 AI-enabled drones by 2027.
Where the hardware is heading, doctrine is scrambling to catch up. Canadian officials told CBC that warfare must stay "a fundamentally human endeavour," and a startup, Dominion Dynamics, says it is building a human-last-decision boundary into its Scout combat drone, yet its own engineers concede they have not worked out what the specific safeguards will be. Craig Martin, an armed-conflict law scholar, warns of automation bias, where operators with seconds to act trust the machine too readily, and of systems failing when used in conditions they were not trained for.
Chinese military writers studying Russia's war in Ukraine have reached a related conclusion: according to a Jamestown Foundation analysis, they judge AI's decisive value to lie in shortening the kill chain, not in fully autonomous weapons. They credit Ukraine's targeting apps with cutting the time from fire order to strike from 20 minutes to 30 seconds, and the People's Liberation Army is building AI planning systems around those lessons while admitting its own doctrine lags the technology.
The diplomatic test comes this month. Diplomats meet in Geneva to decide whether to open negotiations on rules governing lethal autonomous weapons, a conversation Martin and others want Canada to lead. The question on the table is the one the engineers cannot yet answer: what counts as meaningful human control over a weapon that can find and strike a target on its own. The carrier trials show the capability arriving faster than the rules meant to bound it.