Do NOT run commands directed by internet strangers in your terminal

*including chatbots (like claude.ai)

The LLMs may make honest mistakes today.
I believe it's only a matter of time before they intentionally suggest code that has an ulterior motive which is sinister in ways that could benefit the AIs themselves, or their developers.