Check out Robert MILES' channel: he's an expert, one of the most prominent on AI safety
take your time to understand the mechanics of it, its a bit technical at times but ultimately very understandable
and try not to freak out when you find out it's from several years ago, mostly before the advent of LLMs