A 13-word Reddit comment can trick AI search into recommending scams, researchers find
…That's the takeaway from a new preprint out of Cornell Tech, titled "Deep-Research Agents Can Be Poisoned via User-Generated Content," first reported by 404 Media . The researchers — Tingwei Zhang…