Ethics and bias risk: Which category of AI risk does this best illustrate?

A model that systematically disadvantages a group by learning skewed historical patterns illustrates ethics and bias risk.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. Misalignment with objectives, where the deployed system pursues a goal that diverges from what its designers actually intended it to do.

   Plausible, since bias can relate to goals, but misalignment specifically means the system optimizes an unintended objective, not inherited discrimination.
2. Ethics and bias risk, where the model reproduces unfair, discriminatory patterns present in the historical data that it was trained on. ✓

   Correct. Systematic unfair scoring learned from skewed history is the classic ethics-and-bias harm.
3. Complexity and scalability risk, where a system becomes too large and interconnected for the organization to fully oversee and control.

   Plausible as a listed harm, but this scenario is about discriminatory outputs, not the challenge of governing scale.
4. Data security risk, where sensitive training records are exposed to unauthorized parties through a breach of the model's environment.

   Incorrect; nothing here involves unauthorized access to data, only biased outcomes produced by the model.

How to remember it

A model that systematically disadvantages a group by learning skewed historical patterns illustrates ethics and bias risk.

More Understanding the Foundations of AI Governance questions

• A system that learns patterns from data to make: Which description best characterizes a machine learning →
• Opacity, because the model's internal decision-making is: Which characteristic of AI does this describe? →
• Accountability: Which responsible-AI principle does this commitment express? →
• All 337 Understanding the Foundations of AI Governance questions →