liveBook · Manning

Input risk

As explained in AI Governance

Input risk is the risk that arises when user-submitted text reaches a language model.

Input risk arises whenever user-submitted text reaches a language model. It is present even in a basic FAQ bot; additional capabilities can cause other risk categories to accumulate alongside it. 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 ClaimAssist versions to see input risk present from the FAQ bot onward, while data, action, and autonomy risks appear as capabilities expand. section 4.2.1

How AI Governance covers it

How deep it goes
Defines the triggering condition and marks its presence across ClaimAssist versions.
Examples
Uses an FAQ bot and a personalized assistant scenario where a compromised session requests a consolidated customer dossier.
What you can do after
Evaluate whether user-submitted text reaches the model and which additional risks accompany the system’s capabilities.
How it connects
Distinguishes input risk from environment and model risk, while showing data, ability-to-act, and autonomy risks accumulating with added capabilities.

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.