What to Test in AI-Based Systems Micro-Credential

AT*SQA Software Testing Micro-Credential

---

What to Test in AI-Based Systems Micro-Credential

This micro-credential shows you can identify what actually needs testing attention in an AI-based system. That means understanding the risks that are unique to AI, the data privacy and security concerns that come with it, and the quality characteristics that do not exist in traditional software. Passing adds you to the Official U.S. List of Certified Testers and counts toward the AT*SQA AI for Testers certification.

---

Learn What to Test in AI-Based Systems through AT*Learn Training

AT*SQA What to Test in AI-Based Systems Body of Knowledge (Syllabus)

Register for the What to Test in AI-Based Systems Micro-Credential Exam

What the What to Test in AI-Based Systems Micro-Credential Covers

Testing an AI-based system starts with knowing what to look for. This credential is organized around three areas where AI creates testing challenges that traditional software does not.

The first is risk management. AI systems can produce hallucinations, which are outputs that are incorrect but presented as correct. They can make reasoning errors, applying the wrong logic even when individual facts are accurate. They can produce biased results when training data is unbalanced or the algorithm is misconfigured. And because AI is non-deterministic, the same inputs can produce different outputs each time, which complicates how testers define and verify expected results. Drift adds another layer: as a model continues to learn in production, its outputs can degrade over time without obvious warning.

The second area is data privacy and security. AI systems consume large amounts of training data that may include sensitive or personally identifiable information. A pre-trained model may have been built on data whose origins are unknown. The syllabus covers the specific privacy risks this creates, including unintentional data exposure in outputs, lack of user consent over data usage, and compliance exposure under regulations like GDPR. On the security side, it covers AI-specific attack vectors: data exfiltration, request manipulation, data poisoning, and malicious code generation, as well as the mitigation controls testers should understand.

The third area is AI-specific quality characteristics. These are properties that AI systems can have or fail to have, and that traditional software quality frameworks do not account for. The syllabus covers flexibility and adaptability, autonomy, evolution, bias, ethics, side effects and reward hacking, transparency, interpretability, explainability, and safety. Each of these creates distinct testing challenges, and most of them require human judgment from the tester because acceptance criteria are rarely defined in testable terms.

What You Will Learn in the What to Test in AI-Based Systems Micro-Credential

• The causes and symptoms of hallucinations, reasoning errors, and bias in AI systems
• Methods for testing for each, including cross-verification, expert review, consistency checks, logic validation, and output testing
• How non-deterministic behavior affects defining and verifying expected results, including implications for test automation
• How drift works and why AI systems require ongoing monitoring after deployment
• Data privacy risks in AI systems, including unintentional exposure, lack of consent controls, and compliance risk
• AI security attack vectors and practical mitigation strategies: data minimization, anonymization, secure data handling, and securing the model environment
• AI-specific quality characteristics: flexibility and adaptability, autonomy, evolution, bias, ethics, side effects and reward hacking, transparency, interpretability, explainability, and safety
• Why testing AI systems requires judgment from the tester and cannot rely solely on predefined acceptance criteria

Who Should Earn the What to Test in AI-Based Systems Micro-Credential

Software testers and QA professionals working on teams that build or deploy AI-based systems, and anyone preparing for the AT*SQA AI for Testers certification. The AI Introduction for Testers micro-credential is useful background but not required.

How to Earn This Credential

Download the free syllabus to study, then register for the exam at atsqa.org/educational-resources/start for $39. If you want structured video instruction and exam prep materials, the AT*Learn What to Test in AI-Based Systems training is available for $49.

Passing the exam adds you to the Official U.S. List of Certified and Credentialed Software Testers, which employers use to verify credentials. This credential is the second of four micro-credentials in the AT*SQA AI for Testers certification. The others are AI Introduction for Testers, How to Test AI-Based Systems, and Testing Using AI. AT*SQA also offers the ISTQB AI Testing and ISTQB Testing with Generative AI certifications for testers who want a broader AI testing certification.

What to Test in AI-Based Systems: Common Questions

Is this credential harder than the AI Introduction for Testers micro-credential?
The two credentials cover different ground. AI Introduction for Testers covers foundational concepts. This credential focuses on applying those concepts to identify what needs testing in an AI system. It requires understanding specific risk types and quality characteristics, so familiarity with AI basics is helpful, but the exam format is the same: ten questions based on the free syllabus.

Why does testing AI systems require more judgment than testing traditional software?
Traditional software testing relies on defined expected outputs. AI systems are non-deterministic, meaning the same input can produce different outputs each time. Quality characteristics like fairness, transparency, and ethical behavior also rarely come with testable acceptance criteria. The tester has to apply judgment to evaluate whether outputs are correct and whether the system is behaving as intended.

Does this credential cover AI bias testing?
Yes. Bias is covered as both a risk management topic and a quality characteristic. The syllabus covers the two main types of bias, algorithmic and sample bias, how each enters the system, and testing approaches for detecting and monitoring it.

Does AT*SQA offer other AI testing credentials?
Yes. This is one of four micro-credentials in the AT*SQA AI for Testers certification. The others cover AI fundamentals, how to test AI-based systems, and using AI as a testing tool. AT*SQA also offers the ISTQB AI Testing and ISTQB Testing with Generative AI certifications.