It's actually very different. An AI is not something that you just go build and then it's immediately useful. One an AI is built it needs to be trained in a process where it is exposed to examples of real works (text, image, video, etc) and "learns" how to reproduce similar works. This learning process involves using the real works to fine tune parameters of a large mathematical model. These tuned parameters are what make the AI an AI and not just a useless pile of random numbers. So arguably, the owner of every work used to train the model is as much a contributor to the AI as the AI company itself.
These media companies not only want to use artists' works to train a model that they will not be given any ownership of, they want to also own the rights to an artist's image and likeness to be able to reproduce that image and likeness in a potentially infinite number of derived works without needing to compensate the artist.
We investigated ourselves and found we did nothing wrong.
But in all seriousness, this feels like the chronological feed given to users was just a straw man intended to make their algorithm look better. Also the claim in the article that Meta's algorithm does not impact users politics is completely ridiculous and not at all supported by a study like this.