AI-900 Practice Question: Describe features of generative AI workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Fairness

  Why it's wrong here

  Fairness is a responsible AI principle focused on identifying and mitigating bias so that model outcomes are equitable across race, gender, age, and other protected attributes. Prompt injection, however, is an adversarial input technique that tries to override the model's system instructions, which is a robustness and safety concern rather than a bias concern. Since the attack does not target demographic parity or equitable treatment, Fairness is not the relevant principle for this scenario.

• ✓

  Reliability and Safety

  Why this is correct

  Reliability and Safety is the correct principle because Azure OpenAI systems must be trustworthy and behave predictably even when confronted with malicious or unexpected inputs. Prompt injection attempts to subvert the model's intended instructions and can cause the chatbot to output harmful, unintended, or policy-violating content. This principle ensures the system is resilient against such adversarial manipulations, including having safeguards like system messages, input filtering, and abuse detection to maintain safe operation.

• ✗

  Privacy and Security

  Why it's wrong here

  Privacy and Security centers on protecting data confidentiality, controlling unauthorized access, and preventing data breaches through encryption, authentication, and network security. While a successful prompt injection could lead to data leakage, the fundamental issue is that the model's behavior is being manipulated to disregard its configured instructions, which is a failure of reliable and safe operation. The principle of Privacy and Security addresses perimeter and access protections, not the model's susceptibility to adversarial prompt engineering.

• ✗

  Inclusiveness

  Why it's wrong here

  Inclusiveness is a responsible AI principle aimed at designing systems that work well for users of all abilities, backgrounds, and contexts, including accessibility and avoiding exclusion. Prompt injection has no direct relationship to inclusive design; it is an adversarial attack that exploits the model's instruction-following mechanics. Because the attack targets behavioral robustness rather than how fairly or accessibly the system serves diverse users, Inclusiveness is not the applicable principle here.