The premise is wrong. LLMs: only contain information that was generated by humans.

There is no independent intelligence. Its a pattern extraction algo that has some random settings to jump to nearest neighbor nodes.

The constant impulse of humans to attribute agency to it, probably represents its greatest threat.