The negative consequences I see from LLMs mostly come when decision‑makers start prompts with “Is it a good idea to…?”.
Of course, the model replies sycophantically and then spins a positive justification.
Remember: LLMs follow the asker’s framing and favor pleasing replies over facts.