30 experts warn AI is eroding our ability to think and judge

A landmark paper co-authored by 30 AI researchers, including prominent figures like Yoshua Bengio, systematically examines how AI threatens our ability to think and judge. The study, titled 'AI Epistemic Risks: Emerging Mechanisms & Evidence,' identifies three primary mechanisms: persuasion and manipulation (AI systems can exploit human biases for political/economic gain, incite radicalization, or cause unintentional harms like AI sycophancy and mental health risks); cognitive offloading (humans are delegating deeper thinking to AI, risking long-term degradation of individual and societal cognitive resilience); and feedback loops (human-AI and AI-AI interactions narrow the epistemic space, driving homogenization and potentially leading to irreversible fragmentation or 'lock-in').

The experts argue that while AI could boost how humanity processes knowledge, positive outcomes are not guaranteed. Epistemic risks are self-perpetuating—they erode the cognitive and social foundations needed to recognize and govern other threats, including risks from AI itself. The paper proposes actionable solutions across AI system design, human-AI interaction design, institutional adaptation, and information market incentives. The authors stress that the time to act is now, before our capacity to respond is lost entirely.