ACorrect
Infringement of personal information and privacy.
Correct. When personal information is included in prompts or training data and is handled improperly or leaked, it becomes a legal and ethical risk of privacy infringement.
A company is classifying risks by their nature in order to establish an internal policy for using generative AI. Which of the following are appropriate as legal and ethical risks associated with the use of generative AI? (Choose TWO.)
1 / 1
Select all that apply
CorrectA, B
Explanation
A question about choosing TWO legal and ethical risks of generative AI.
- 1「appropriate as legal and ethical risks」Choose issues related to law and ethics
- 2「(Choose TWO.)」Privacy infringement and biased output apply
BCorrect
Unfairness and loss of trust from biased or discriminatory output.
Correct. When a model containing bias produces biased or discriminatory output, it leads to unfair treatment and loss of trust in the company, which is a legal and ethical risk.
CIncorrect
A system being manipulated through prompt injection.
Prompt injection is a security threat that overrides the model's instructions with crafted input and makes it behave unexpectedly.
It is a serious risk that requires mitigation, but it falls under security risk rather than the legal and ethical risk category this question asks about, so this is incorrect.
DIncorrect
Prediction accuracy gradually declining due to model drift.
Drift is a quality and operational matter where prediction accuracy gradually drops as the input data trend shifts from training time.
It is a risk to address with monitoring and retraining, but it falls under quality and operational risk rather than the legal and ethical risk category, so this is incorrect.
EIncorrect
Migration becoming difficult due to dependence on a specific vendor.
Dependence on a specific model or platform is a strategic and procurement risk (vendor lock-in) that makes it harder to migrate to another option later.
It is a point worth considering, but it falls under business and strategic risk rather than the legal and ethical risk category, so this is incorrect.
Key Takeaway
Generative AI risks can be classified by their nature. Infringement of personal information and privacy and unfairness and loss of trust from biased or discriminatory output are legal and ethical risks related to law, non-discrimination, and transparency (others include intellectual property infringement and the spread of misinformation). On the other hand, prompt injection (security), model drift (quality and operations), and vendor lock-in (strategy and business) are all real risks but have a different classification axis.