This NATO-backed drone AI picked its own target and struck it, but relied on
a human-drawn map to do so

Date:
Mon, 21 Sep 2026 21:05:00 +0000

Humans assigned the drone with its mission, but no one has explained whether it can be stopped if its links are jammed -- A small drone flying over a snow-covered Swedish test range in January 2026 found several objects, ranked an armored engineering vehicle as the most valuable, and made a beeline for
it before dropping an explosive.

There was no human involvement in the entire process, which included selection, targeting, and then neutralization by the drone, making it a considerably more autonomous approach than what the current battlefield uses. The flight, part of BAE Systems Bofors' Affordable Loitering Modular Ammunition (ALMA) program with onboard AI supplied by Uppsala-based Scaleout Systems, resurfaced this month as AI safety became a key topic among industry and government stakeholders alike. Scaleout Systems tested
the AI in question with a pilot on standby and a continuous
connection throughout the exercise, during which the drone used AI to detect, identify, and geolocate potential threats.

Depending on how you frame it, you can argue that this is both a threat to modern militaries and a boon: an autonomous drone that can select targets without oversight or a backlink is essentially immune to jamming, offering a much higher likelihood of completing its mission than alternatives.

At the same time, it becomes riskier when it is completely cut off from communications, malfunctions, or fails to select the correct target without human oversight.

Interestingly, BAE Systems Bofors' own release on Winter Demo 2026 does not mention the strike at all. It frames the three-day Karlskoga event around roughly 20 startups and technology companies, with Swedish Defense Minister
Pl Jonson joining a panel and saying, "We need more actors and entrepreneurs in the defense industry."

The test, despite lacking human interaction in its end-game, still relied on human input to some degree: a user pushed a button to start the exercise, and humans set the parameters overseeing the entire exercise.

This isn't the first machine to choose its own target, however; Israel's
Harpy loitering munition, fielded since the 1990s, can autonomously detect
and attack radar emitters whose locations were unknown at launch.

This is awkward timing for regulators, to say the least: on September 5, the 128 states party to the Convention on Certain Conventional Weapons agreed on
a non-binding text on lethal autonomous weapons, after the United States and Russia pushed late changes , including to a section on human judgment and control.

As a result, Nicole van Rooijen of Stop Killer Robots told Reuters that three years of work had been "substantially diluted in the last hours." For now, Scaleout has shown that detection, ranking, and targeting can run on a small airframe without a data link, within parameters that people set.

Whether this makes it to a modern battlefield is increasingly likely, even as the Russia-Ukraine conflict becomes more reliant on autonomous machines. The more pressing question is one that will be answered sooner or later: how does such a drone make decisions if it is jammed from its source without any human oversight?

Link to news story: https://www.techradar.com/pro/this-nato-backed-drone-ai-picked-its-own-target- and-struck-it-but-relied-on-a-human-drawn-map-to-do-so

$$
--- MultiMail/DOS
* Origin: capitolcityonline.net * KY, USA (1:2320/107)