> LLMs have more potential for weaponising personal information against us.

Doesn't follow from

> They will comply with anything requested in the prompt.

LLMs are impressive in how they are able to produce plausible sentences, but they don't actually bring any new data correlating or collecting capabilities to the table.