There is a difference between being full of shit and plausible deniability. Their statement is as follows:
"We pledge that we will not weaponize our advanced-mobility general-purpose robots or the software we develop that enables advanced robotics and we will not support others to do so. When possible, we will carefully review our customers’ intended applications to avoid potential weaponization."
Which in layman's terms is:
"We don't SUPPORT groups weaponizing the robots and we won't give them special software to do so buuuuuut if they buy a robot for search and rescue purposes (and put that down for their use-case) only to later strap rifles to it? Well that's on them for using it against our wishes. We can't help what people do with their property"
My friend works in a open cast mine in Eastern Europe. The miners like to drink but it is strictly forbidden. They use mining explosives and one day decide if would be funny to blow up a stray dog. They strap the explosives to the dog and retreat to cover. About to trigger the explosive and turn round to find the dog standing next to them.
One of the scary things to me about this is that if something that can be made to look like this robot becomes available in a cheap enough model (same applies for drones) you can seek to overwhelm an enemy by swarming them with mostly harmless even cheaper versions (think toy robots and drones) and force the enemy to 3aste resources taking all of them out knowing that some of them might be armed, but not which.
E.g. you can buy $4k or so quadruped robots on Aliexpress while Boston Dynamics' ones reportedly costs in the $200k range. If you get that kind of ratio, for twice cost you can "augment" your armed robot with 50x decoys to drive up your opponents cost to eliminate the threat 50-fold..
In other words, I expect when these end up getting used, sooner or later it will be swarms of them...
Why are people in the comments talking like this is the beginning of an AI uprising? This thing is still remotely controlled by a soldier and not autonomous. It's about as dangerous as any gun, which is to say, pretty dangerous but nothing we haven't seen before. If you want an army of weaponised ribo dogs you'd also need an actual army of people to control them
Largely, I expect, because of the point you make. Needing an actual army of people to control them becomes a limiting factor. Add on to that that requiring remote control makes them vulnerable to jamming, and there's a strong incentive to start making them more and more autonomous both to enable fewer soldiers to control more bots, and to allow them to retain some function without it.
It just largely seems like it will be too significant a temptation.