A few things to note based on the comments here.

(1) This isn't particularly new or noteworthy. This kind of thing has been done for at least a decade. They claim better results than previous efforts, but their examples don't appear categorically better. Setting aside previous efforts, the results here are just not that good. They kinda get a sense of what the participants are viewing, but that's it.

(2) This isn't mind-reading in the colloquial sense that people are interpreting it as. They are using brain activity while participants are actually viewing images, not while they are imagining them. That is a big difference and is much easier than anything that would generally be considered "mind-reading".

(3) Even if it was mind-reading, and even if it actually was high-fidelity, this requires a million dollar MRI machine and having a participant basically bolted onto a sled for a couple hours. All the comments by people talking about how we're all doomed and privacy is gone seem to be missing that fact.