Artificial Intelligence Centrality in Psychotic Delusions and Violence Risk in Forensic Psychiatry: Retrospective Observational Study of Judicial Decisions

Background: Artificial intelligence (AI)–themed delusions are increasingly observed in psychotic-spectrum disorders, reflecting the incorporation of contemporary sociotechnical elements into delusional systems. However, it remains unclear whether the structural role of AI within these belief systems is associated with increased violence risk or more restrictive forensic outcomes. Given the importance of dynamic clinical factors (eg, insight and treatment adherence) in forensic risk assessment, clarifying the role of AI centrality has clinical and legal relevance. Objective: This study examined whether AI centrality within psychotic delusional systems is associated with (1) violence toward others and (2) judicial findings of significant public safety risk in forensic psychiatric decisions. Methods: This retrospective observational study used jurisprudential data from the Société québécoise d’information juridique database, including all publicly available Quebec tribunal and court decisions up to December 31, 2025. Eligible cases (N=29) involved psychotic-spectrum disorders with explicit AI-related delusional content and judicial consideration of dangerousness or disposition. The unit of analysis was the judicial decision. AI centrality was coded as central (n=15, 51.7%) or noncentral (n=14, 48.3%) using a structured, text-based framework. The primary outcome was documented violence toward others; secondary outcomes included direct AI-linked violence attribution and judicial findings of significant public safety risk. Covariates included impaired insight, treatment nonadherence, substance use disorder, and prior violence history. Data were extracted through full-text review using a standardized coding grid. Bivariate associations were analyzed using Fisher exact tests (α=.05), and odds ratios (ORs) with 95% CIs were calculated. Exploratory logistic regression models were performed to assess adjusted associations. Results: Violence toward others was documented in 20/29 (69%) cases. AI centrality was not significantly associated with violence (12/15, 80.0%, vs 8/14, 57.1%; OR 2.91, 95% CI 0.63-13.45; P=.26) but was strongly associated with direct AI-linked violence attribution (9/15, 60.0%, vs 2/14, 14.3%; OR 9.00, 95% CI 1.48-54.6; P=.01). Judicial findings of significant public safety risk were more frequent in AI-central cases (13/15, 86.7%, vs 9/14, 64.3%; OR 3.60, 95% CI 0.63-20.5; P=.24), although not statistically significant. AI-central cases demonstrated higher prevalence of impaired insight (13/15, 86.7%, vs 8/14, 57.1%; OR 4.89, 95% CI 0.79-30.1) and treatment nonadherence (9/15, 60.0%, vs 4/14, 28.6%; OR 3.75, 95% CI 0.74-18.9). Conclusions: AI centrality within delusional systems appears to be not independently associated with increased violence toward others but is strongly associated with AI-based attribution of behavior and markers of epistemic vulnerability, including impaired insight and treatment nonadherence. The findings suggest that AI-themed delusions function as structural organizers of meaning and agency rather than novel criminogenic risk factors. Clinically and legally, this underscores the importance of prioritizing dynamic risk variables over thematic novelty, informing more proportionate forensic decision-making and risk assessment in an era of rapidly evolving digital environments.