Dark patterns are everywhere online. As LLM-based web agents are increasingly used for online tasks, it’s crucial to ask how these agents handle them. Our recent work, accepted to IEEE S&P'26, shows agents are susceptible at significant rates. Preprint: https://t.co/YRHkq0jeNQ