Ex-GCHQ chief: AI could give drones a workable moral code

TL;DR:

• Former GCHQ director David Omand says AI could now build a “moral” framework letting autonomous weapons distinguish combatants from civilians.
• It reverses his 2014 view that such systems could not comply with international humanitarian law.
• Campaigners call the position “as nonsensical as it is dangerous”.

A former head of GCHQ has changed his mind on one of the most contested questions in modern warfare. David Omand told the Guardian that generative AI now makes it possible to give unmanned weapons a workable moral framework — more than a decade after he chaired a commission that doubted autonomous drones could ever meet the laws of war.

From “in the loop” to “on the loop”

Omand argues the speed of modern conflict — drones, hypersonic missiles — is making real-time human authorisation of every strike impractical. The answer, he says, is shifting from humans “in the loop” (approving each action) to “on the loop” (setting parameters and supervising). He proposes an “adaptive moral control layer” in which a commander sets variables before a mission, such as the acceptable proportion of civilians near a target, encoding existing military ethics rather than inventing new ones. The result, he contends, could even be “ethically superior” to split-second human judgement.

The context is an accelerating shift towards machine-led targeting. UK armed forces minister Al Carns has said there may be circumstances requiring “the ability to take the human out of the loop”, while the US 2027 budget earmarks $54bn (£40bn) for autonomous systems and a “drone dominance” programme. AI from Palantir and Anthropic has already been used to shorten the “kill chain” in the Iran conflict — echoing how Iran has turned Western AI models into cyberwar tools.

Looking forward

Not everyone is persuaded. Chris Cole of Drone Wars UK said AI “merely processes data”, lacking any real ability to judge proportionality or distinguish civilians. That objection goes to the heart of the debate: whether moral reasoning can be parameterised at all. Omand — now a King’s College London professor and defence-sector adviser — is calling for urgent work on the question, arguing the worse outcome is sleepwalking into AI weapons with no moral component built in at all.