Well, there are a lot of threats from AI that are popularised today. Some of them are fake, crazy, or stupid, but others are real. Here are some threats that are real, in my opinion.

In trying to use AI for good, we may give too much power to an AI that we do not understand well enough, and whose motivations are not clear. Worse still, this AI might have motivations that we humans could not understand, even if we wanted to. Such an AI may not value humans at all.

Another threat is that a country like China creates an AI and puts it in charge of their soft power program. It may be possible for an AI to be so intelligent that it could manipulate the world in ways that humans physically could not understand. This could allow China to not only literally take over the world, but also ensure that everybody with any kind of power is absolutely thrilled about it. (Anybody without the power to stop this AI would be disregarded in its calculations, no matter how much they hate the AI. This could easily include most humans.)

One thing AI has already been doing for a decade to our great detriment is optimising the ad revenue of search engines and social networks, with a total disregard for all other consequences.

All of these threats are not new. People have been talking about them for years, now. You should pay more attention.