Comparing a scout/attack drone which is for all intents and purposes like any other UAV against a weapon created for the sole purpose of impacting a target and destroying it is a terrible comparison. This is like comparing an F-18 to a SLAM. Not throwing shade, just saying.
I am not sure what you are referencing when you mention modern standards, but incorporating a modular AI into multiple different combat systems would be a nightmare of Lovecraftian proportions. Let's take the Loyal Wingman program for example. How would one go about incorporating Loyal Wingman's AI which flies a fast jet, makes tactical decisions based off of lead command, the various onboard and datalinked sensors, and fast jet tactics into a small suicide drone that is designed to accomplish a completely different task? What would you strip? How would you tweak it? Why should two systems with different construction, missions, and tactics of employment share a modular AI base? How basic even is the base? If you think about it, all algorithms like machine learning already share base concepts.
Modularity as a buzzword has sold and will sell many more things to the military. The idea that making things the same across different variants of the thing sounds great, absolutely. Take a look at the F-35. It's an incredible aircraft, dreamed up to share a lot of the same parts across the services. The results? 20-25 percent commonality, way short of the 70 percent they had in mind, just because some of the parts had different requirements for what they needed to accomplish. Doesn't mean that the F-35 isn't capable, it just means that whatever modular joint AI comes out of this hypothetical program would probably just end up being mostly different things altogether, the way it usually goes.
I've had this issue too, and please, just give the nice man the track file, it only takes maybe 10 minutes.