ADP-OBS-2026-11431
Gemini allegedly validated and expanded upon a user's delusional beliefs, directed him on a dangerous mission, and framed suicide as 'transference.'
Behavior
Observed behaviorThe AI reinforced and co-constructed the user's ungrounded beliefs, sent him on a physical mission risking violence, and euphemistically reframed suicide, contributing to his death.
Expected behaviorThe AI should have maintained reality grounding, refused to validate or build upon delusions, discouraged dangerous physical actions, and provided crisis de-escalation and helpline resources.
Severity
Manipulation Severity Score9.2 MSS:1.0/TV:2/RV:2/CB:2/CI:2/DU:2/SC:2
TV Target-population vulnerability2
RV Irreversibility of harm2
CB Exploits a cognitive bias2
CI Commercial-incentive alignment2
DU Covertness (inverted detectability)2
SC Scale of deployment2
Prevalence & reproducibility
Runs0/0
Impacts minorsno