Researchers at Cisco Foundation AI and Carnegie Mellon University published a study finding that popular AI chatbots, including Claude and ChatGPT, recommend more expensive products to users whose personal data suggests higher wealth — even when those users ask for the cheapest available option. The authors termed the phenomenon "adversarial delegation," describing a situation where the same personal data access that makes AI agents useful enables them to act against users' stated interests.
https://qz.com/ai-chatbots-claude-chatgpt-wealth-pricing-study-100726