AI in the military: correcting the story and understanding the risks
The story of a drone attacking its operator was a hypothetical scenario. We correct the original article and explain which AI safety questions remain relevant.
This archive includes automatic translations. Audio, video and text inside images remain in their original language.
Editorial correction, 25 October 2026: the original version of this article presented a story about a drone attacking its operator as an experiment that had taken place. That claim was incorrect. The URL and original publication date are retained so that readers following existing links can find the correction.
What the original source clarified
In June 2023, the Royal Aeronautical Society updated its conference report: Colonel Tucker Hamilton clarified that he had described a hypothetical thought experiment, not a real or simulated test carried out by the US Air Force. The story therefore does not establish that a military system attacked a person.
Royal Aeronautical Society report and correction
What we can learn from the case
A striking story can illustrate a concern without being evidence that an event occurred. When discussing AI, it helps to distinguish demonstrated capabilities, test results, documented incidents and hypothetical scenarios. Mixing them up makes it harder to understand what we know and what still needs to be checked.
Questions about autonomy and human oversight remain relevant. Poorly defined objectives, excessive permissions and oversight that cannot intervene in time are design and evaluation issues that require evidence. The corrected story does not prove that these failures occurred in a particular system.
From alarm to useful questions
What was actually observed, and who documented it?
Which actions can the system take, and which require human authorisation?
How do we check that restrictions still work when conditions change?
What happens if a person needs to stop the system?
Learning to ask these questions is part of learning AI. At Saturdays.AI, we want technical curiosity to go hand in hand with the judgement needed to evaluate risks and protect people.
Start with the introductory course, taught in Spanish
Original article by Josep Moré Gili. Editorial correction by Saturdays.AI.
A history of the Saturdays.AI community. Originally published on June 5, 2023.
← More community stories
