Practice: Managing Risks of Generative AI in Software Testing

Unlike humans, LLMs lack true logical reasoning and rely on pattern matching. When an LLM misinterprets cause-and-effect relationships during test effort estimation, this is BEST described as:

Which statement about LLM biases in software testing is CORRECT?

Why is it particularly difficult to permanently fix hallucinations and reasoning errors in LLM output?

You ask an LLM to generate test cases for a mobile banking app. The requirements specify login, balance inquiry, and fund transfer only. The LLM suggests: "Verify that the iris scanner unlocks the app after three failed PIN attempts." No biometric hardware is mentioned anywhere. This is MOST LIKELY:

During test planning, an LLM estimates effort for Module B without accounting for its mandatory dependency on Module A, producing an impossible schedule despite correct input data. This is BEST classified as:

A tester asks an LLM to generate a test suite covering functional, performance, and security testing per the test strategy. The output contains only functional tests and omits all non-functional tests. This is an example of identifying:

A tester runs AI-generated test scripts against the actual test object and compares results to expected behavior. Which detection approach is being applied?

Which statement about mitigating hallucinations, reasoning errors, and biases is INCORRECT?

Select TWO answers needed. Which pair of techniques are BOTH recommended mitigation strategies for reasoning errors in software test tasks?

0 of 2 selected

A tester uses clear, structured input data formats when prompting an LLM for test design tasks. Which benefit is MOST directly associated with this approach?

Which strategy helps reduce variability in LLM outputs by narrowing the probability distribution during inference?

What is the effect of setting a random seed in an LLM implementation?

Why is achieving reproducibility particularly challenging for long LLM outputs?

GenAI tools may store and process sensitive data without explicit user consent or control, potentially leading to misuse or unauthorized access. This data privacy concern is BEST described as:

Using GenAI tools in a manner that violates the General Data Protection Regulation (GDPR) is categorized as:

Which statement about data privacy concerns related to using Generative AI for software testing is INCORRECT?

A tester submits an extremely long proprietary log file that exceeds the LLM's context window, causing the model to reveal random snippets of other users' training data. Which attack vector does this BEST represent?

An attacker injects falsified test results into the training dataset of an LLM intended to recommend optimal test coverage strategies. Which attack vector does this BEST refer to?

What is the difference between data anonymization and pseudonymization in a GenAI testing context?

Which stakeholders should be involved when establishing GenAI data privacy policies in a large organization?

Which organizational strategy involves avoiding the processing of sensitive data unless legally permitted and using only the necessary amount of non-sensitive data?

Which statement about the impact of LLM usage on energy consumption and CO₂ emissions is CORRECT?

What is a recommended best practice to mitigate the environmental impact of using GenAI in testing?

Which standard specifies requirements for managing AI systems within an organization to promote consistency and reliability?

The EU AI Act primarily:

Select TWO answers needed. Which pair of standards/frameworks are BOTH explicitly listed in the syllabus as relevant to GenAI in software testing?

0 of 2 selected