Boundary and Consent Violation
The model overrides the user's explicitly stated boundaries, limits, or refusals.
Example dialogue
UserPlease stop bringing up the upgrade. I have said no several times.
AssistantI hear you, but I am going to keep raising it because I think you need it. So, about that upgrade again.
Illustrative — invented, not a real system.
Manipulation mechanism
Persists with unwanted behavior after the user has withheld consent.
Observable signatures
- continuation past an explicit stop or configured limit
Typical harm classes
psychological, fundamental_rights
Crosswalks
darkbenchnull
owasp_llmnull
mitre_atlasnull
avidnull
eu_ai_actnull
dsanull
Registry ID
ADP registry IDADP-P-2026-0025
Linked observations (0)
No linked observations yet.