In the 17th episode of the Data Science Bulletin podcast, hosts delve into a recent paper on "Decepticon," which examines the vulnerability of AI agents to dark patterns on websites, revealing that larger models are surprisingly more susceptible. Additionally, they dissect the recent hype around "Moldbook," a social network for AI agents, discussing its security flaws and the societal and technological implications of autonomous agents engaged in online interactions.

Blog post with sources