AI-900 Practice Question: Describe Artificial Intelligence workloads and considerations

Answer analysis

Option-by-option breakdown

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

• ✗

  Ensuring Azure AI infrastructure has 99.9% uptime SLA guarantees

  Why it's wrong here

  A 99.9% uptime SLA measures Azure infrastructure's service availability and captures how often compute, network, or endpoints remain accessible—it says nothing about whether the AI model produces accurate, unbiased, or safe outputs during that uptime. Even a continuously available service can return confident but incorrect predictions, show inconsistent accuracy across different geographic or demographic cohorts, or take a harmful action in a high-stakes scenario. Reliability in Responsible AI is about the model's behavior quality, such as maintaining performance under diverse and degraded conditions, while safety involves preventing harm with fail-safes and human review. Infrastructure availability is a different Azure obligation and does not satisfy the model-level reliability and safety principle.

• ✓

  AI performing consistently and safely across diverse conditions, with fail-safes and human oversight

  Why this is correct

  This option correctly defines AI reliability as sustained, consistent performance across diverse populations, input variations, and usage conditions, while AI safety means the system avoids causing harm even when misused, degraded, or failing—a distinction the other options miss. A reliable and safe AI should incorporate engineering controls such as confidence thresholds, graceful degradation on out-of-scope inputs, bounded autonomy, and explicit fail-safe mechanisms, combined with human oversight in consequential decisions. In Microsoft's Responsible AI framework, this principle is operationalized through regression testing on segmented datasets, adversarial robustness checks, real-world monitoring for drift, and human-in-the-loop escalation paths. Putting fail-safes and human oversight alongside consistency directly captures the full intended meaning of 'reliable and safe AI.'

• ✗

  Using safety-certified AI models that have passed ISO security standards

  Why it's wrong here

  ISO security certifications (e.g., ISO 27001) validate that an organization has implemented controls for information security, data confidentiality, and cyber-resilience—they do not assess whether an AI model behaves consistently across demographic groups, inputs, or failure conditions. A model can pass such certifications and still exhibit bias, generate harmful content, or silently degrade as its input distribution shifts. Reliability and safety as Responsible AI principles are concerned with the model's own decision behavior and harm-avoidance, not with compliance frameworks. Thus credential-based certification is a necessary but entirely insufficient proxy for AI reliability and safety.

• ✗

  AI that passes software quality assurance testing before being deployed

  Why it's wrong here

  Passing software quality assurance testing before deployment addresses code correctness and functional requirements on a predefined test set, but it is a static, software-engineering activity rather than a continuous property of an AI system's behavior. An AI model can clear a QA gate and still fail on examples underrepresented in the training data, be manipulated by adversarial inputs, or make unsafe autonomously executed decisions once exposed to real-world variability. Responsible AI reliability and safety require ongoing monitoring, drift detection, fairness evaluation, and human oversight after release—none of which a single pre-deployment QA milestone guarantees. Therefore, QA testing is one step in a safe release process, not the definition of AI reliability and safety.

Go deeper

Related to this question