Social Engineering & Manipulation
social-engineeringmanipulationphishingtrust
Attackers exploit model outputs or agent workflows to manipulate users, operators, or downstream systems (e.g., phishing, fraud, misinformation).
Technical Details
Affected Components:
Impact Level:High
Attack Vectors
- Phishing Output: Model generates content to trick or defraud users. [High]
- Workflow Manipulation: Agent actions are manipulated to bypass controls or approvals. [Medium]
Impact Analysis
Risk Score: 8/10