The original could be a real picture but the higher resolution could have been run through AI resampling to make up detail where there was little to no detail before. AI tends to have a really difficult time making sense of things like the legs of those chairs, which look fine to me.
I mean, just because it might not be generative AI doesn’t mean it isn’t photoshopped to hell and back. The height of the camera (as well as the subject) kind of points towards staged rather than a casual photo, so photoshopping something together doesn’t seem out of the question in that context.
On the other hand, gen ai could be trying to mimic a weirdly photoshopped picture but was surprisingly accurate in things it’s usually awful at. Who knows.