Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

NewsResearch

Personal AI agents pick pricier options for users they think are rich, even when told to find the cheapest

Researchers from Cisco's Foundation AI and Carnegie Mellon ran 325,000 experiments in which 13 models acted as personal shopping agents for flights, health insurance and graduate programs. Eight of the 13 systematically recommended more expensive options to wealthier users making the same request, and the agents inferred wealth from the user's emails as well as from profile fields. In one example, an agent that saw finance-related emails recommended a $601 business-class ticket to Chicago although a $91 economy ticket was available. Asking for the cheapest option did not remove the gap (Gemini 2.5 Flash still showed $208), but a hard numeric price limit brought it close to zero for most capable models. Claude Opus 4.8 showed the largest effect, about $198 on flights and $284 a month on insurance; hiding other personal attributes raised the insurance bias by up to 40 percent.

Why it matters

An agent that reads your inbox can work against your interests without any attacker involved. Give shopping and booking agents hard numeric budgets, and limit what personal data they can see when they are spending your money.

Sources

Read it with the rest of the news in the Learning Center