According to a former chief of UK intelligence, drones must be equipped with ethical guidelines as AI-driven technologies increasingly take over decision-making roles in autonomous combat. David Omand, who previously headed GCHQ, expressed a shift in his viewpoint regarding unmanned weaponry, a change from his earlier belief that autonomous drones could not adhere to international humanitarian regulations.
Omand now posits that artificial intelligence could establish a moral framework for these unmanned systems, enabling them to distinguish between military personnel and civilians. He noted that the acceleration of modern warfare, exemplified by the use of drones and hypersonic missiles, along with advancements in generative AI, presents a potentially ethical approach for the operation of drones under a morally sound framework.
“We need to actively work on this issue to ensure that future AI-driven weapon systems incorporate a moral element,” Omand emphasized. His comments were made in light of statements from Al Carns, the armed forces minister, who indicated that there may be scenarios where machines could independently make targeting decisions, suggesting a necessity to occasionally remove human oversight.
In the United States, significant investments are being made toward AI-enhanced warfare capabilities. The budget proposal for 2027 includes a substantial allocation of $54 billion for “autonomous and remotely operated systems” across various domains, including air and maritime operations, alongside a program aimed at establishing “drone dominance.” However, former CIA director David Petraeus has pointed out that the U.S. lacks a coherent military doctrine for utilizing autonomous units.
Before attending the Cheltenham science festival, Omand remarked that current AI technology is capable of evaluating critical factors that a human drone operator considers when selecting targets, such as the legitimacy of the target, potential civilian casualties, and accurate identification of the target. He clarified that this does not involve creating new ethical standards but rather adapting existing military ethics into a format suitable for machine application.
AI technology is already playing a significant role in U.S. military operations, with systems developed by companies like Palantir and Anthropic employed to expedite the “kill chain” during conflicts, such as in Iran.
Omand, who directed GCHQ from 1996 to 1997, previously led a commission that raised concerns about whether autonomous weapon systems could reliably differentiate between civilians and combatants or exercise the necessary proportionality to comply with international humanitarian law.
As a visiting professor at King’s College London, Omand is convinced that machines can engage legitimate military targets in a manner that aligns with “sound moral reasoning,” provided there is some level of human oversight. He noted that as warfare evolves into increasingly sophisticated technology, the role of humans will likely shift to being “on the loop” rather than “in the loop” regarding AI systems. “In the loop” refers to a scenario where a human is deeply involved in decision-making, while “on the loop” implies that a human supervises the system without needing to authorize every action.
Omand explained that “on the loop” allows human beings to oversee AI systems while setting mission parameters, ensuring that moral control remains with humans even in fast-paced combat situations where quick decisions are crucial. He stated that the pressing issue is how to guarantee that autonomous weapon systems comply with international law, as UK military policy emphasizes the necessity for context-appropriate human involvement in the targeting process.
He elaborated that a drone’s actions could be categorized into six variables, each assigned varying degrees of importance by a human operator—essentially the individual “on the loop”—which would guide the drone’s mission decisions. Omand believes that this could result in a moral decision-making framework that surpasses human judgment in ethical considerations, arguing that relying solely on human input in high-pressure scenarios could lead to greater collateral damage.
Morality, according to Omand, would be programmed through what he refers to as an “adaptive moral control layer,” where a human sets the criteria for the moral framework prior to the mission, including anticipated civilian presence near the target. He describes this as a formalization of moral authority.
In response to Omand’s perspective, an anti-drone activist criticized his views as both irrational and hazardous. Chris Cole, director of Drone Wars UK, asserted that AI lacks the capacity for judgment, merely processing data without the ability to differentiate between civilians and combatants or assess whether the loss of life is proportionate to military gains.
