A trust relationship is an expectation that another entity’s requests or actions are legitimate. In a multi-agent system, that expectation can let hidden instructions travel from one agent to another and trigger access the first agent should not have. section 4.2.1
ClaimAssist and scheduling
An attacker embeds hidden instructions in a claim. ClaimAssist asks the scheduling agent to check availability; trusting the request, that agent follows the instructions and retrieves the adjuster’s full calendar, including other claimants’ visit addresses. section 4.2.1
Table 4.1 How each additional ClaimAssist feature increases the security risk
|
Version
|
Environment
|
Input
|
Model
|
Data
|
Ability to Act
|
Autonomy
|
|
V1: FAQ Bot
|
Yes
|
Yes
|
No
|
No
|
No
|
No
|
|
V2: FAQ Bot
running on an unvetted model
|
Yes
|
Yes
|
Yes
|
No
|
No
|
No
|
|
V3:
Personalized Assistant
|
Yes
|
Yes
|
Yes
|
Yes
|
No
|
No
|
|
V4:
Fine-tuned Assistant
|
Yes
|
Yes
|
Yes
|
Yes
|
No
|
No
|
|
V5:
Action-Enabled Assistant
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
No
|
|
V6: Agentic
Assistant
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Compare how ClaimAssist’s security exposure grows as it moves from a FAQ bot to an agentic assistant with the ability to act. section 4.2.1

