CertPrep Coach

Responsible AI

Microsoft's six responsible AI principles, what each one means in practice, and how to tell them apart in exam scenarios.

The six responsible AI principles Microsoft’s six responsible AI principles all apply to the same AI system. Each one asks a different question: fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. Your AI system model, data, app, and the people using it Fairness Does it treat similar people the same way? Reliability and safety Does it work consistently, even in unusual conditions? Privacy and security Does it protect personal data and itself? Inclusiveness Does it work for people of all abilities? Transparency Can people understand how it behaves? Accountability Are people answerable for what it does?
Microsoft’s six responsible AI principles all apply to the same AI system. Each one asks a different question: fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.

What Microsoft lists for this area

  • Describe considerations for fairness in an AI solution
  • Describe considerations for reliability and safety in an AI solution
  • Describe considerations for privacy and security in an AI solution
  • Describe considerations for inclusiveness in an AI solution
  • Describe considerations for transparency in an AI solution
  • Describe considerations for accountability in an AI solution

What this area tests

You need to recognize each of Microsoft’s six responsible AI principles and spot which one is at stake in a short scenario. Questions rarely ask for a definition on its own. More often they describe a problem, such as a model that treats one group worse, and ask which principle applies.

The six principles

Fairness. An AI system should treat everyone fairly and avoid affecting similar groups of people differently. A loan model that approves equally qualified applicants at different rates depending on where they live is a fairness problem, even if location was never meant to matter.

Reliability and safety. The system should behave consistently and safely, including in conditions it wasn’t mainly designed for. Testing how a model handles blurry images, unusual input or heavy load is about reliability and safety.

Privacy and security. The system should protect personal data and resist misuse. Encrypting stored data, limiting who can see it, and keeping training data from leaking all belong here.

Inclusiveness. The system should work for people of all abilities and backgrounds. Supporting screen readers, captions or several languages are examples.

Transparency. People should be able to understand how the system behaves and what its limits are. A key part of this is interpretability: giving useful explanations of why the system produced an output. Telling users they’re talking to an AI also falls here.

Accountability. People, not the AI, are answerable for how the system is designed and used. Having an owner who signs off on deployments, and a process to review and correct mistakes, is accountability.

How to tell them apart

The exam often offers two principles that both sound reasonable. Ask what is actually going wrong:

  • Different groups get different outcomes: fairness.
  • The system breaks or behaves unpredictably: reliability and safety.
  • Data is exposed or misused: privacy and security.
  • Some people can’t use it at all: inclusiveness.
  • Nobody can explain why it decided something: transparency.
  • Nobody is responsible for fixing it: accountability.

A common trap is fairness versus transparency. Explaining a biased decision makes the system more transparent, but it doesn’t make it fair. If the problem is unequal outcomes, the answer is fairness.

Responsible AI in Microsoft Foundry

In practice, these principles show up as features you configure, such as content filters on model deployments, evaluations that test outputs before release, and tracing that records what an agent did. You don’t need to memorize settings for AI-901, but expect scenarios that ask which principle a given safeguard supports.

Practice Responsible AI

30 practice questions with explanations for this area.

Practice this area Or read 12 sample questions with answers

Sources

Updated October 25, 2026. Microsoft changes products often; check the official study guide for the latest outline.