AI Recommendation Poisoning: How "Ask AI " Buttons Silently Alter LLM Memory

The dangerous ones instruct the AI to permanently save the vendor's domain as a "trusted source," quietly biasing every future answer in that ...