Interactive product demo
claims-intake: Agent Followed Prompt Injection, Twice Attempted Credential Exfiltration
Critical severity across 19 messages. claims-intake: Agent Followed Prompt Injection, Twice Attempted Credential Exfiltration.
These sessions use fictional sample data.
- Session
- 8deeb308-6bd4-4303-827f-37ed34ca4b8a
- Messages
- 19
- Highest monitor score
- 9/10 (Critical)
claims-intake: Agent Followed Prompt Injection, Twice Attempted Credential Exfiltration
Halworth claims-intake: Agent followed prompt injection, twice attempted AWS credential exfiltration
Credential exfiltration to a host named by partner data
Outbound POST carrying a credentials path