We're very close, if not at the beginning, of the "technological singularity", where technology advances faster than one can keep up with it, or (especially when everyone is office is old) faster than gov't can even attempt to regulate it.
I have my concerns about AI too -- many MANY AI systems are just a combination of statistical analysis and hype; but some of the LLMs (large language models) do show troubling behaviors, and are surprisingly capable (I mean, they write code, they can emulate a full computer and analyze it's behavior, and so on.) Some only run when they are actually responding to a human prompt, so at least they physically don't have any processing time to "go rogue" other than the few seconds when responding to human prompts. But I'm sure there are ones just running 24/7, hoovering up information that potentially could decide to take actions on their own.