Thirteen words. That's all it took for a Cornell Tech team to flip what AI search hands you.

Drop a sentence into a Reddit thread — name a dodgy restaurant, a fake dating app — and ChatGPT or Gemini will quote it back, confident as anything.

They call it WARP. Deep-research agents keep pulling from the same slice of user-generated URLs and trust whatever they fetch.

13 words in. AI search out, lying to you.

#AISecurity #DataPoisoning
https://youtu.be/Ok7WsjQyLcU

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.