ISTQB Specialist (CT-GenAI v1.1) Probeprüfung #3

The third full-length practice exam for the ISTQB Certified Tester Testing with Generative AI (CT-GenAI) v1.1 syllabus, in force since April 2026. It follows the official exam format exactly: 40 questions worth 46 points, 60 minutes, and 30 points to pass (65%). Scoring is weighted the way the real exam weights it — K1 and K2 questions are worth 1 point and K3 questions 2 — so the six applied questions carry 12 of the 46 points between them. The chapter split is the official one: 7 questions from Introduction to Generative AI, 11 from Prompt Engineering (16 points, and the home of five of the six K3 items), 10 from Managing (AI) Risks, 5 from LLM-Powered Infrastructure and 7 from Deployment & Integration. The K-level mix matches the published structure as well: 8 questions at K1, 26 at K2 and 6 at K3. Mock exam 1 concentrates on prompting for test analysis, test data and pilot governance, and mock exam 2 on model behaviour, technique-driven derivation and operational risk. This one takes a third angle: making generated work checkable. The first chapter covers the parts of a model's setup that decide what an answer even is — system prompts, output length limits, token cost, model cards, self-reported confidence. The prompting questions ask how to turn a vague request into one whose answer can be verified: a named coverage criterion, a traceability mapping, pairwise coverage over four parameters, a risk-based smoke selection backed by evidence, an oracle that reports its own assumptions, and a load profile derived from a non-functional requirement. The risk chapter deals with what organisations actually have to decide: licence provenance of generated code, a data incident in a public repository, confidentiality classification rather than a single data type, vendor lock-in before scaling, the case against letting one assistant both draft and review, and what a count of generated test cases does and does not prove at a sign-off. Chapter 4 looks at retrieval as something that is operated — a stale index, what a citation establishes, telling a retrieval failure from a generation failure, a regression set over prompt and index changes, and access control enforced at retrieval. Chapter 5 covers the organisational frame: governance group, approved-tool list, exception process, independent review in a safety-related project, the skills the work now needs, and what makes a result transfer to a second team. Every option, correct and incorrect alike, carries its own rationale, and every question has an overall explanation. As with the other two mocks, remember that CT-GenAI is about testing with generative AI; testing AI-based systems is the subject of CT-AI, which is a separate exam.