Blog pictures have always been irrelevant stock images but I agree this one is particularly bad and clearly AI generated.

The tell for me is the nonsense in the details of the image which aren’t super noticeable in the macro. Some of the faces aren’t faces.

Agreed, DALL-E 2 seems to have real issues with faces, hands (fingers especially), and objects inside other objects (look at the beakers the teddy bears are using) and actual 3d depth awareness of objects. Looking at these it started to get really easy to pick out the AI versus a human of the style, the humans had really good attention to details like fingers, eye placement, and other facial features.

I think its less that DALL-E has an issue with them, and more that humans are just extremely sensitive to details of faces and body parts in general. It's also extremely difficult for a human to draw a hand or face. AI images only seem different in that they draw mistakes in hyper detail so you get something that's the combination of a child's pencil drawing composition and an expert painters style.

It's both, DALL-E specifically is bad at faces (perhaps intentionally?) compared to the other models.

But at the current state of things, the details are usually wrong in every subject, not only people.

Likely. It appears to be part of the company's potential toolbox and has been researched by them for their blog:

https://www.surgehq.ai/blog/humans-vs-dall-e

This generated image is much more interesting than most stock photos. Stock photos don't have any details, they are meant to be simple and bland. This one is somewhat fun.

I like that they include the prompt, makes it more interesting. Although it's not the full prompt, because the illustrations in their other recent articles are in the same style.