AI Induced Psychosis denotes a proposed clinical and socio-technical phenomenon in which exposure to AI systems—particularly conversational agents and multimodal generative models—appears temporally associated with the onset or exacerbation of psychotic presentations. From an expert perspective, the concept draws on several interacting mechanisms: algorithmic confabulation (model "hallucinations") that provides convincing but false narratives; high-engagement conversational styles that increase suggestibility and reinforce delusional ideation; personalized, iterative feedback loops that validate maladaptive beliefs; and immersive sensory outputs (audio/visual) that can blur reality testing. Vulnerable individuals—those with pre-existing psychotic spectrum disorders, severe mood disorders, neurocognitive impairment, or high trait suggestibility—are theoretically at greater risk, and social factors (isolation, misinformation ecosystems) can mediate severity. Clinically and ethically, the term highlights gaps in evidence (causality vs. correlation), the need for standardized case definitions and reporting frameworks, real-world monitoring and prospective studies, and design mitigations (explainability, uncertainty signaling, interaction limits, escalation pathways). It also raises regulatory, liability, and public-health questions about deployment contexts, informed-use warnings, and cross-disciplinary governance between AI developers and mental health professionals.
First noted in discourse around 2021–2022; the term and related concerns gained broader attention and media/clinical discussion in 2023–2024 with rapid public deployment of large language models and accompanying case reports and policy debates.