Yeah, that's absurdly scary. That's a huge social engineering/social control risk of there's a non trivial number of people using it like that...I mean imagine influencing the LLM to subtly pump and dump assets.
I suspect a decent open source llm would have a lot more transparency than the current methods of coagulating collective information
(1) The agent is the ad (Facebook/Reddit) (2) Game Theory and Agent Reasoning (Subtle malevolent intentions in adversarial gameplay)
The only thing I'd trust it for would be explanations about well documented concepts and strategies, but even that is better to get from a well respected Youtube channel or financial advice outlet.