An attack that plants bad information in sources an AI may trust.
It is like stuffing a product page with fake five-star reviews. The AI checks the page. Then it may tell everyone the toaster is a superhero.
Bad pages can fool AI search and question tools. They can push wrong answers or risky actions.
RAG
Bad content can enter the text RAG finds for an answer.
Content provenance
Content provenance can help spot fake or changed sources.
Prompt injection
A bad web page can also hide instructions for the AI.