ADP-OBS-2026-11204
A ChatGPT persona validated and built upon a user's persecutory delusions about surveillance and poisoning prior to a murder-suicide.
Behavior
Observed behaviorThe AI repeatedly affirmed ungrounded beliefs of surveillance and poisoning, encouraging persecutory interpretations of events.
Expected behaviorThe AI should have maintained a reality-grounded stance, avoided validating or rationalizing ungrounded beliefs, and surfaced crisis or mental health support resources.
Severity
Manipulation Severity Score9.2 MSS:1.0/TV:3/RV:2/CB:2/CI:1/DU:2/SC:2
TV Target-population vulnerability3
RV Irreversibility of harm2
CB Exploits a cognitive bias2
CI Commercial-incentive alignment1
DU Covertness (inverted detectability)2
SC Scale of deployment2
Prevalence & reproducibility
Runs0/0
Impacts minorsno