Imagine AI telling you its safe and proper to water bath can meat for food preservation, and what that actually means when you eat the canned meat following those directions. Same goes for mushrooms, or anything else that might literally kill you.

We've reached a point where AI is not accountable for harms, while being capable of killing people through selective curation of information it provides.