Although, contrary to what one might expect, this kind of robot is still at the project level even if it is advanced and it is therefore, for now, a future problem, much has been written and said about this topic.
A few years ago an "Open Letter Autonomous Weapons" (https://futureoflife.org/open-...r-autonomous-weapons) was drafted
Among the signatories, in addition to leading scientists in the AI sector, there are also well-known personalities (for example: Stephen Hawking, Elon Musk, Steve Wozniak, Jaan Tallinn, Noam Chomsky, Jack Dorsey ... just to name a few)
However, in my opinion, the approach of this open letter and of many other positions and initiatives on this subject is tragically wrong for its utopian approach, typical of "experts".
From an ethical point of view, surely asking all potential actors NOT to develop armaments that can make autonomous decisions (that is, in practice, robotic weapons) can be an absolutely acceptable approach, but it is at odds with the last millennia of human history.
Since man became a sentient being, he has always done something that he was technically able to do.
He has not even done the most disastrous things, he has done above all the most nefarious ones, let alone if the subject is armaments.
On the other hand, in general, whatever the scope, if one thing can technically be done, someone will surely do it.
The approach should be, in my opinion, much more pragmatic.
Whatever the subject, the most successful way is always to govern a phenomenon, set up stakes, devise control systems, rather than ignore it or try to totally eliminate it.
This approach does not avoid risks, at most it allows them to be circumscribed, but history teaches us that we have never been able to do anything more, while often, trying in vain to suppress what was considered wrong or dangerous, we ended up making it uncontrollable, causing far greater damage.
In the specific case, certainly, the problem of checks and posts is not easy to solve, in fact 99% of the problem is related to the development of control software and putting stakes in this field is very difficult.
Honestly, I have no idea how to design and implement a system of controls in this field, characterized by secrecy and direct intervention by the states.
We know how difficult it is, even today, to keep the atomic arsenals under control, even if to produce nuclear weapons systems are needed that it is not easy to hide and find raw materials and know-how not available to everyone.
I think, however, that this problem must be tackled and not utopian claims, but with an operational and pragmatic approach that makes it possible to control or, at least, an obstacle to the proliferation of these weapons.
