The linked PDF (Storm-2035 [1]) from Microsoft is more detailed and interesting than the blog post. However, what's missing from the reports is how they detected those operations and how they tied them to different groups. There's a lot of claims being made without showing all of the supporting evidence.
To give them the benefit of the doubt, they likely want to keep their detection methods secret to make circumvention more difficult. And it all sounds totally plausible of course. But at the same time, a degree of skepticism is warranted because Microsoft has a huge incentive to fearmonger about AI so they can lock it down and capture the market. And what better way is there than to use the usual bogeymen to do so.
To give them the benefit of the doubt, they likely want to keep their detection methods secret to make circumvention more difficult. And it all sounds totally plausible of course. But at the same time, a degree of skepticism is warranted because Microsoft has a huge incentive to fearmonger about AI so they can lock it down and capture the market. And what better way is there than to use the usual bogeymen to do so.
[1] https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcor...