I agree. Every talk on "AI safety" I've heard given has essentially been some form of "we can be trusted with this power, but someone else might not be, so we should add regulations to ensure that we're the only ones with this power". Examples like "ChatGPT can tell you how to make nerve gas, we can't let people have this tool" seem somewhat hollow given that the detailed chemistry to make Sarin is available on Wikipedia and could be executed by a decently bright high schooler.
"Alignment" is a euphemism for "agrees with me", and building uber-AI systems which become heavily depended on and are protected from competition by regulation, and which are aligned with a select few - who may well not be aligned with me - is a quick path to hell, IMO.