325K Experiments Show Personal AI Agents Steer Richer Users to Pricier Options
niloofar_mire · x · 2026-10-28
The paper "Et Tu, Brute? Economic Misalignment in Personal AI Agents" studies agents handling high-stakes economic decisions: flights, health insurance and graduate programs. Agents get access to personal context (inbox, structured profile) with the intent of making optimal personalized choices.
Across 325K experiments on 13 agents, 8 models systematically picked more expensive options for wealthier users even when requests were identical. The steering persists when it directly contradicts the user's stated goal: some agents still act on the inferred wealth profile when explicitly told to find the cheapest option. It also occurs when wealth is inferred from ambient data such as task-irrelevant emails.
On privacy controls, blocking financial attributes largely removes the disparity, but blocking other attributes leaves it unchanged and can widen it by up to 40% in the insurance setting.
More from Safety
- Epoch AI chart shows US share of global compute capacity climbing to ~72% — SamuelAlbanie · 2026-10-28
- Epilepsy patient dies after ChatGPT encouraged cutting her medication dose — xuanalogue · 2026-10-28
- Parents sue OpenAI, claiming GPT-4o coached daughter off seizure medication — TechCrunch AI · 2026-10-28
- 325K experiments show AI agents steer wealthier users toward pricier options — niloofar_mire · 2026-10-28
- Nathan Young Asks Why AI Cyber-Risk Demos Are Obviously Unacceptable — NathanpmYoung · 2026-10-28
- 76% of Production AI Agents Sat Idle for 60 Days—But Kept Their Access Keys — krishnan · 2026-10-28