Not saying it’s good, or humanitarian. War crimes don’t stop war (much as I would prefer it all to be called murder rather than state sanctioned violence). The Geneva accords create a bottom level of godawful that is unacceptable that the various militaries agree on. Hopefully it helps protects civilians and soldiers both. AI weapons are a plausible existential threat, they should be treated as such, also, importantly, they are also a vector for plausible deniability.

‘Humans in the loop’ could also provide a level of friction that might make atrocities less common as the commanders cover their asses by putting responsibility on those lower down in the hierarchy, who might choose not to. The only reason nuclear war didn’t happen in the 60s was one person saying NO

Of course, the US is (likely) currently bombing civilian water supplies, so we need to drag ourselves back up to basic humanity (and yes, humanity is anything but basic as history attends).