The technology (at least with current methodologies) is flawed: that's why people are warning of the bubble bursting. We can't properly scale LLMs on our current grid in the same capacity as China. Our technologies are also incredibly energy-intensive compared to their technologies.
There is no intelligence, the hallucinations are likely fundamental, the cases of people being given dangerous or harmful advice are rising, human AI psychosis is a real concern, the sycophancy/bias confirmation is still present, and major actors in the AI space are existentially afraid of any form of regulation of the technology/industry (which does not signal confidence).
Also, it's critical to factor in the whole copyright issue with training data... one domino is all it takes to collapse the whole thing.