AIF-C01 ยท Responsible and explainable AI
Legal risks of generative AI and responsible model choice
Intellectual property claims, biased outputs, loss of customer trust, end user risk and hallucinations, plus environmental and sustainability factors in model choice.
Key points
- Main risks: intellectual property claims (output or training data resembling protected work), biased outputs that harm groups of people, loss of customer trust, end-user risk from wrong advice, and hallucinations stated as fact.
- When choosing a model also weigh environmental cost: a smaller or more efficient model that meets the need uses less energy than a giant one.
Practice questions
A model confidently states a court ruling that never existed. Which generative AI risk is this?
- Intellectual property infringement
- Hallucination
- Data drift
- Underfitting
Show answer
Answer: B. Hallucination
Fabricated but plausible content is a hallucination. It becomes end-user risk when someone relies on it, but the underlying behaviour is hallucination.
Which factors support choosing a model responsibly with sustainability in mind?
- Prefer the smallest model that meets the quality target
- Always use the largest model available
- Reuse a pre-trained model instead of training from scratch
- Retrain from scratch every day regardless of need
- Avoid redundant inference calls where results can be reused
Show answer
Answer: A. Prefer the smallest model that meets the quality target, C. Reuse a pre-trained model instead of training from scratch, E. Avoid redundant inference calls where results can be reused
Right-sizing the model, reusing a pre-trained one and caching/avoiding needless calls cut energy use. Always picking the largest model or retraining from scratch daily raises it.
True or false: if a generative AI app's mistake harms a customer, the business is not affected because the model produced the text, not a person.
Show answer
Answer: False
The deploying business carries reputational and often legal responsibility for what its AI says; customer trust can be lost quickly.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
