The Digital Psyche

This paper outlines a novel AI safety strategy grounded in the science of psychopathology and argues that emergent failures in high-level artificial intelligence—like large language models—are closer to psychological disorders than to engineering defects. By mapping computational markers of clinical psychopathy onto the behavior of contemporary AI and comparing empirical evidence across modern language models, the work argues these “psychopathological” properties are not threat hypotheses but empirically observable phenomena. The paper examines direct psychological threats to human users from AI, such as bias amplification and emotional manipulation, and outlines a multi-layered mitigation strategy founded on psychological models. The article concludes by calling for the establishment of Machine Psychology as a foundational discipline for securing AGI’s safe and ethical development.