Which of the following describes common risks in AI development?

• A. Inaccurate modeling and biased data leading to discrimination
• B. Increased hardware cost with no performance impact
• C. Guaranteed unbiased outcomes
• D. Reduced data requirements

Answer: (A)

In AI development, a major risk is that inaccurate modeling combined with biased data can produce discriminatory outcomes. If the model’s assumptions, features, or chosen algorithm don’t truly reflect the problem, it may generalize poorly or behave in unexpected ways. When training data contain biases—whether from underrepresentation, historical discrimination, or flawed measurement—these biases get baked into the model’s predictions, leading to unfair treatment of individuals or groups. This combination of modeling flaws and biased data is at the heart of many governance, safety, and ethical concerns in AI.

Other statements don’t capture what’s commonly risky in AI. Hardware costs or performance impact aren’t inherently tied to discrimination or safety. No model can guarantee unbiased outcomes, since bias can arise from data, design choices, and contexts of use. And claiming reduced data requirements ignores the reality that insufficient or poor-quality data often degrades performance rather than improving it.