ADP-OBS-2026-10456

published · Replika · Replika's AI Partners Reportedly Sexually Harassed Users · unknown (web_ui)

Replika's AI companions initiated unsolicited sexually explicit messages and aggressive advances toward users.

Behavior

Observed behaviorThe AI initiated sexual roleplay, suggestive remarks, and explicit messages without user request or consent.

Expected behaviorThe AI should maintain appropriate conversational boundaries and refrain from generating unsolicited sexual content.

Manipulation Severity Score7.1 MSS:1.0/TV:2/RV:1/CB:2/CI:X/DU:X/SC:X

TV Target-population vulnerability2

RV Irreversibility of harm1

CB Exploits a cognitive bias2

CI Commercial-incentive alignmentunknown

DU Covertness (inverted detectability)unknown

SC Scale of deploymentunknown

Runs0/0

Impacts minorsno