ISTQB CT-AI Certification Exam Sample Questions

CT-AI Dumps PDF, AI Testing Dumps, download Artificial Intelligence Tester free Dumps, ISTQB AI Testing exam questions, free online Artificial Intelligence Tester exam questionsYou have to pass the CT-AI exam to receive the certification from ISTQB. To increase the effectiveness of your study and make you familiar with the actual exam pattern, we have prepared this ISTQB AI Testing sample questions. Our Sample ISTQB Certified Tester AI Testing Practice Exam will give you more insight about both the type and the difficulty level of the questions on the ISTQB Artificial Intelligence Tester exam.

However, we are strongly recommending practice with our Premium ISTQB Certified Tester AI Testing (CT-AI) Practice Exam to achieve the best score in your actual ISTQB CT-AI Exam. The premium practice exam questions are more comprehensive, exam oriented, scenario-based and exact match of ISTQB Certified Tester AI Testing exam questions.

ISTQB AI Testing Sample Questions:

01. A biometric access-control system guards a hospital's controlled-drug store. Measured over all enrolled staff, it meets the error thresholds agreed for it. A review then reports that false rejections are several times more frequent for one group of staff than for another, so those staff are repeatedly turned away at the door and have to call a supervisor across to let them in.
Which AI-specific quality characteristic from ISO/IEC 25059 does this finding bear on most directly?
Select ONE answer.
a) Societal and ethical risk mitigation - a quality-in-use characteristic under freedom from risk, which covers harm the system does to a group of people even when its overall error figures are acceptable.
b) AI robustness - recognition holds up when presentation conditions are poor.
c) Intervenability - a supervisor can override a rejection and open the store.
d) AI functional correctness - a product-quality characteristic for which an acceptable error threshold is agreed, since perfect accuracy is unachievable, and this system sits inside its threshold.
 
02. A contractor's model for detecting missing personal protective equipment was built with a widely used, mature machine learning development framework. On that basis a project manager proposes to reduce the planned system testing of the safety-monitoring feature.
Which response is correct?
Select ONE answer.
a) The proposal is sound, because a mature framework's components are tested, documented and widely used, so the defects system testing looks for are already removed.
b) The proposal is unsound, because the framework supplies building blocks, not the data or the acceptance criteria.
c) The proposal is sound for the model, though the application built around it still needs system testing because it was hand-written.
d) The proposal is unsound, because a framework's built-in components stay unverified until the project has tested them itself.
 
03. A flood-warning service is built from two components. One grades river-level risk with fuzzy logic, using membership functions and rules elicited from hydrologists. The other predicts runoff with a neural network fitted to gauge and rainfall records. A tester who is new to the project asks for the training dataset of each component so that a data-quality review can be run on both.
Why can that request be met for only one of the two components?
Select ONE answer.
a) Fuzzy logic reuses the neural network's dataset, so reviewing that one dataset covers both components at once.
b) Fuzzy logic is fitted to unlabelled sensor data rather than to labelled examples, so a dataset does exist but a conventional data-quality review cannot be applied to it.
c) Fuzzy logic is a conventional technique, so it falls outside the AI part of the system and outside the scope of AI-specific testing.
d) Fuzzy logic is an AI technology that encodes expert knowledge as membership functions and rules, so its behaviour comes from that encoded knowledge and there is no training dataset to review.
 
04. A housing association answers tenant enquiries with an assistant built on a supplier's pretrained large language model. Between two test cycles the supplier's model version is unchanged and no fine-tuning has been carried out, but the operations team has edited the standing instruction that is sent ahead of every enquiry, adding a paragraph on repair timescales. Testers re-run their recorded enquiries, get materially different answers, and are told that nothing in the AI has changed.
Which statement about that situation is correct?
Select ONE answer.
a) An added paragraph can only affect an enquiry that asks about repair timescales, so the other answers changed for an unrelated reason.
b) Editing the standing instruction amounts to fine-tuning, so the model has in effect been retrained and its version number should have risen.
c) Identical weights must return identical answers for a recorded enquiry, so the difference comes from the sampling settings.
d) The standing instruction forms part of the input the model is conditioned on, so editing it changes the assistant's behaviour just as a new model version would.
 
05. A manufacturer wants machine translation for the manuals of products that have not yet been announced. Company policy forbids sending unreleased product information to any external party, and the team has no capacity to train a translation model from scratch.
Which option satisfies both constraints, and why?
Select ONE answer.
a) Train a translation model in-house from the company's own archive, because only a model trained on company data can respect the policy.
b) Use a public hosted translation service, because manual text is technical documentation, and technical documentation is not commercially sensitive material.
c) Use a public hosted service to draft the translation and have staff redact the unreleased detail afterwards.
d) Deploy a pretrained translation model on infrastructure the company controls, because no training run of its own is then needed and the manual text never leaves the company.
 
06. A museum is adding two machine learning components to its cataloguing system. The first assigns each photographed artefact to one of forty predefined material categories. The second drafts a short descriptive paragraph about the artefact for the public gallery guide.
Which statement correctly characterises these two components?
Select ONE answer.
a) Only the second component is generative AI, because its output is new text, not a predefined category.
b) Neither component is generative AI, because generative AI is present only where a large language model is involved.
c) Both components are generative AI, because each one produces a category, a paragraph or a value that the catalogue did not previously contain.
d) Only the first component is generative AI, because it infers a material attribute that no catalogue record states.
 
07. A rail operator's signalling assistance system has two components. A conventional interlocking component applies the rulebook that the safety engineers wrote, refusing any route that violates a coded condition. A newer component recommends headway adjustments; it was built by fitting a model to five years of recorded train movements, and the engineers who built it cannot point to a rule inside it that produces any particular recommendation.
Which property makes the second component AI-based rather than conventional?
Select ONE answer.
a) It reaches its recommendation faster than a human dispatcher could.
b) It must be re-verified after every change, whereas the interlocking component is verified once at commissioning.
c) It runs on dedicated accelerator hardware, whereas the interlocking component runs on general-purpose controllers.
d) Its behaviour was fitted to recorded movements, so no inspectable rule accounts for a given recommendation.
 
08. A shipping operator plans to add route optimisation to its fleet-management product by consuming a supplier's pretrained model through a hosted interface, rather than building and hosting a model itself.
Which TWO of the following are consequences for the test team of that decision?
Select TWO answers.
a) The supplier may change the model behind the interface, so observations recorded today may not be reproducible afterwards.
b) Testing is limited to what can be observed from inputs and outputs, because the training data and model internals are unavailable.
c) The team can retrain the supplier's model on its own accumulated voyage history whenever measured functional performance falls below the agreed target level.
d) Acceptance criteria for route quality pass to the supplier, since the supplier owns and operates the model.
e) The team must provide AI-specific processors, network capacity and storage of its own, so that inference throughput and latency match what the fleet requires.
 
09. An acoustic anomaly detector listens to a compressor hall and flags bearing faults. When the plant replaces two of the compressors, the background noise of the hall changes permanently. The specification requires the detector to go on flagging faults correctly under the new noise floor by re-fitting its own baseline in service, without the supplier shipping a new model.
Which AI-specific quality characteristic from ISO/IEC 25059 does that requirement express?
Select ONE answer.
a) Flexibility - the system can be reconfigured to run under operating conditions beyond those specified when it was first built.
b) AI robustness - functional correctness is held despite noisy, distorted or adversarial input to the model.
c) User controllability - an operator can retune the detector's baseline when hall conditions change.
d) Functional adaptability - the system autonomously adjusts to changes in its operational environment after deployment.
 
10. An insurer is agreeing acceptance criteria for a model that extracts fields from claim documents. Two of the criteria read as follows: the extraction must stay correct when scans are skewed, faint or stamped across the text; and every claimant must be told which fields on their claim were filled in by the model and what those values were based on.
Which TWO of the AI-specific quality characteristics defined in ISO/IEC 25059 do these two criteria address?
Select TWO answers.
a) Evolution
b) AI robustness
c) Interpretability
d) Flexibility
e) Transparency

Answers:

Question: 01
Answer: d		 Question: 02
Answer: b		 Question: 03
Answer: d		 Question: 04
Answer: d		 Question: 05
Answer: d
Question: 06
Answer: a		 Question: 07
Answer: d		 Question: 08
Answer: a, b		 Question: 09
Answer: d		 Question: 10
Answer: b, e

If you find any errors or typos in ISTQB Certified Tester AI Testing (CT-AI) sample question-answers or online ISTQB Artificial Intelligence Tester practice exam, please report them to us on feedback@processexam.com

Your rating: None Rating: 4.7 / 5 (91 votes)