The issue is mainly in regards to carbon footprint and climate change. As far as I'm aware, it's thankfully getting significantly better (EDIT - arguably it hasn't, the training process still uses way too much energy across the board even if the consumer's consumption has improved) - especially in regards to imaging - and varies from company to company, but most generative AI training technology, especially those that are currently widely available - ChatGPT, (though focused on text is still generative), is arguably the most infamous example - uses
a lot of energy thanks to the intense computing power required. This is mainly in regards to training the system rather than individual consumer use.
I'm certainly not an expert on this as my stance on AI relates more to art than to scientific and computer jargon I don't fully understand, but here's an
article that explains it leagues better than I could!
Also for those who don't understand the issue of AI in the art world, in a briefly explained nutshell AI "art" doesn't make its own images from scratch - it is trained off of existing images - often stolen by AI companies from artists who did not give permission - which are fed into a machine to be stitched together into something "new".