Those bananas are completely different. There's no copyright infringement there. I could take a photo of a banana and photoshop it repeatedly onto a pink background. That would look just as similar, and there's no copyright problem there.
The images are obviously different, but it appears that DALL-E maps the getty images description to similar tone, similar perspective, similar background, and similar weather conditions. I'm sure there are thousands of possible backdrops in Guangzhou, and many ways to show a parkour flip. Even in the Google image search results there's more variance than in the output of DALL-E.
So you can't copyright an idea, but you can certainly scrape a copyrighted DB with image metadata, and use it to create your own product. My point is that DALL-E itself might be a derivative work of Getty Images and thousands of other online catalogs.
Interesting. Adding "stock photo" to the string generated that getty tag? That is probably the most attackable (alas easy to fix) part of the issue. It will be an interesting question how close to the original a picture has to be to be considered the same (I'm sure there's some case law) and maybe there's some new research to be done regarding how to recreate the training data images with the correct search string (I suppose one could build an ML model for that).
I tried it with a distinctive banana image:
https://imgur.com/a/0OrIr6e