The US airforce is testing versions of future AI-controlled armed drones to be used in air combat alongside human fighter pilots.

I hope the article errs in saying that this AI is similar to the erroneously labeled "AI" that operates today's bullshit generator language models, because if that is accurate we have no telling who it might fire at.

One of the dangers of AI-controlled devices that can do things with real physical effects — even if not armed — is that it may make clever inventions that could be fatal in ways that the AI would not understand, and that no human would be asked to check. This AI is already demonstrating that sort of creativity which is also potential danger. I recommend reading the book The Two Faces of Tomorrow, by James Hogan. See libgen.rs.