I'm sorry if this isn't worded well, the idea isn't yet crystallized in my mind such that I can precisely express it. (foreshadowing)

Basically, when the 'Al' is given a prompt, it takes a random image of chaos and adjusts the variations in the pixels over many steps into a target or goal image.

Youtube - Computerphile - Stable Diffusion In Code

Pretty awesome. The video I linked he made frogs on stilts. Wild.

But what if the parameters of this algorithm are changed? Say, the input goal instead of an image of frogs you ask for Complete Control of Earth. Then, instead of starting with random noise, it starts with a snapshot of the entire world in its current state. The Al adjusts the virtual world in small increments, revealing a path of little steps from point A to point B.

Whoever runs this algorithm learns of a way to turn chaotic Earth under his complete control. All he needs is for all the world to stop and smile for the photo. How long did it take to Stable Diffuse a plan for world domination? About two more weeks...

In conclusion, the "AI art" Stable Diffusion algorithm can be used as a plan-maker. Think of who gathers the most algorithm-training data and who has the most computational power.

P.s. if they're giving free versions of this tech to the general public, does that mean they have applied the algorithm and are already enacting their plan? Is the Villain revealing his scheme because he feels we're inescapably trapped?