It's known as "criti-hype".

Google:

• Criti-hype critics: We have to rein Google in, they spy on absolutely everything you do, and as a result their advertising is like a mind control ray!
• Google: Hey advertiser, did you hear, even our critics say our advertising is like a mind control ray, that's why you have to pay us so much.
• The reality: Google doesn't even know enough to avoid showing people ads in languages they don't understand despite all their settings and all their usage being in a different language.

Palantir

• Criti-hype critics: Palantir tech must be stopped is so powerful that a human isn't even needed in the decision loop!
• Palantir: Hey government, did you hear, even our critics say that we don't need a human in the decision loop!
• The reality: Palantir suggests bombing a school.

We're basically already at a point where most enterprise documentation is written partly or entirely by AI, and then read mostly by AI. Eventually true understanding of e.g. how to build a docker-compose file will be genuinely rare. And they want to apply that model to everything.

The risk isn't that GAI will decide, in an Asmovian sense, that the best way to make peace on earth is to kill us all, or that to achieve it's manufacturing quota it must transmute all matter in the universe I to paperclips. The real risk is that the MBA-wielding professional-managerial class will not be able to resist the siren song of demolishing labor with something that works "well enough." Eventually they're going to put it in charge of something critical like ATC or dam control or the electrical grid.

Humans, by our nature, are diverse, and one fuck up rarely cascades through the entire global system. We make small fuck ups and train each other. AI "learns" by observing and probabilistically imitating, and one model could be deployed in a thousand places and screw up the exact same way.

tldr ai won't kill us, our bosses will once Anthropic wines and dines them

Maybe it'll just kill the billionaires?