It's not that controversial in full blown war zones. Air defense systems already have to operate autonomously for instance, since not every threat will be dete

It's not that controversial in full blown war zones. Air defense systems already have to operate autonomously for instance, since not every threat will be detected early enough to allow for human review. Plus, friendly fire incidents also happen with humans, we don't punish those unless they're obviously negligent, just work on ways to not repeat that mistake, so how is AI any different? The determination of negligence would just fall upon whoever was responsible for configuring the drone for that mission. So, for example, if the drone was used to target a certain kind of tank which is in use on both sides and it was known ahead of time that the drone can't differentiate between them, it'd be negligent use, while a mission programmed with full understanding of the drone's capabilities results in friendly fire due to a previously unknown error, it'd be no different from any other current smart weapon messing up.