liveBook · Manning

Risk factor

As explained in AI Governance

A risk factor is a dimension of a generative AI system’s exposure that can increase its likelihood or consequences of failure or exploitation.

A risk factor is a dimension of a generative AI system’s exposure that can increase the likelihood or consequences of failure or exploitation. Exposure grows with what a system can see, do, and decide; each additional factor compounds risk. section 4.2

Two factors illustrate how exposure can arise from the surrounding system as well as the model:

  • Environment: weak logging, secret management, access controls, vendor security, or unvetted plugins can expose an application. An insecure database can create serious risk even when the model is not at fault. section 4.2
  • Model: reliance on an unvetted or uncontrolled model can introduce risks from malicious training instructions, unassessed weights, vendor updates, or fine-tuning that poisons data, weakens safeguards, or memorizes sensitive information. section 4.2
Figure 4.2 Security Risk Factors. Each additional factor compounds the risk.

The figure summarizes how additional risk factors compound exposure. section 4.2

How AI Governance covers it

How deep it goes
Defines exposure and details environment and model factors within a six-factor framework.
Examples
Uses weak infrastructure controls, risky model sourcing or fine-tuning, and an exposed database containing chat histories and API keys.
What you can do after
Evaluate an application’s exposure by checking its environment and model trust, control, and fine-tuning risks.
How it connects
Connects risk to security controls, infrastructure, models, fine-tuning, vector databases, orchestration libraries, and plugins.

Related concepts

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.