Free SecAI+ CY0-001 Practice Test

12 free SecAI+ practice questions

1. Q1Basic AI Concepts Related to Cybersecurity

   Question 1. A web security team has years of labeled examples of benign and malicious URLs. The team wants a tool that learns from those examples and predicts whether newly observed URLs are likely to be malicious. Which of the following BEST describes the AI approach being used?

   1. ATransformers
   2. BGenerative adversarial networks (GANs)
   3. CGenerative AI
   4. DMachine learning

   Show the answer

   Correct answer: D. Machine learning

   Machine learning uses data to learn patterns and make predictions, such as classifying new URLs based on prior examples.

2. Q2Securing AI Systems

   Question 2. A security architect is reviewing a new customer support chatbot that uses a large language model (LLM) and retrieval-augmented generation (RAG). The architect needs a concise, industry-recognized resource that focuses on common LLM application security risks to use during the design review. Which of the following BEST meets this need?

   1. AOpen Worldwide Application Security Project (OWASP) Machine Learning Security Top 10
   2. BCommon Vulnerabilities and Exposures (CVE) AI Working Group
   3. COpen Worldwide Application Security Project (OWASP) Top 10: LLM Top 10
   4. DMassachusetts Institute of Technology (MIT) AI Risk Repository

   Show the answer

   Correct answer: C. Open Worldwide Application Security Project (OWASP) Top 10: LLM Top 10

   The OWASP LLM Top 10 is the best fit because the system is an LLM-based application and the team needs a focused list of common LLM security risks.

3. Q3AI-assisted Security

   Question 3. A security operations engineer maintains detection-as-code rules in a local editor. The engineer wants AI assistance to flag malformed rule logic, inconsistent field names, and style issues before the rules are submitted for peer review. Which of the following BEST meets this requirement?

   1. AA personal assistant that reminds reviewers when rule approvals are overdue
   2. BAn integrated development environment (IDE) plug-in for code quality and linting of detection rules
   3. CA browser plug-in that evaluates suspicious websites for phishing indicators
   4. DA chatbot that generates a high-level summary of the organization’s security strategy

   Show the answer

   Correct answer: B. An integrated development environment (IDE) plug-in for code quality and linting of detection rules

   An IDE plug-in is best because the engineer is editing detection-as-code locally and needs inline code quality and linting assistance.

4. Q4AI Governance, Risk, and Compliance

   Question 4. A public university is preparing datasets for several artificial intelligence (AI) research projects. Faculty members need someone to ingest records from student information systems, normalize fields, maintain data pipelines, and ensure the project teams receive well-structured datasets for modeling. Which of the following roles is the BEST fit?

   1. AData engineer
   2. BAI risk analyst
   3. CAI governance engineer
   4. DMachine learning engineer

   Show the answer

   Correct answer: A. Data engineer

   A data engineer prepares and manages data pipelines and structured datasets that other AI roles use for modeling and analysis.

5. Q5Basic AI Concepts Related to Cybersecurity

   Question 5. A cybersecurity startup trained a phishing detection model. The model performs extremely well on the training messages but performs poorly on a separate set of newly collected test messages. Which of the following BEST describes what model validation has revealed?

   1. AThe model requires no further evaluation because training performance is high.
   2. BThe model is likely overfitting to the training data.
   3. CThe model has completed quantization successfully.
   4. DThe model is using unsupervised learning correctly.

   Show the answer

   Correct answer: B. The model is likely overfitting to the training data.

   The validation result suggests overfitting because the model learned the training data too specifically and does not generalize well to new messages.

6. Q6Securing AI Systems

   Question 6. A legal research vendor is preparing a large language model (LLM) feature that drafts case summaries for attorneys. The vendor added a guardrail requiring the model to avoid unsupported citations and to state when a source does not support a conclusion. Which of the following is the BEST way to validate the guardrail before release?

   1. AReduce the model’s output length so summaries are less likely to include multiple citations.
   2. BUse only the model vendor’s general benchmark score to determine whether the guardrail is ready for production.
   3. CAsk a few attorneys to use the feature informally for a week and report any summaries that seem incorrect.
   4. DCreate a guardrail test suite with supported, unsupported, ambiguous, and adversarial citation scenarios, then verify expected responses before deployment.

   Show the answer

   Correct answer: D. Create a guardrail test suite with supported, unsupported, ambiguous, and adversarial citation scenarios, then verify expected responses before deployment.

   A dedicated guardrail test suite validates whether the model behaves correctly across normal, edge, and adversarial cases.

7. Q7AI-assisted Security

   Question 7. A security operations center (SOC) analyst at a global retailer receives several suspicious customer-service emails written in different languages. The analyst needs to quickly understand the content, identify the likely intent, and brief the incident commander in English. Which of the following AI-enabled security use cases BEST supports this task?

   1. ASignature matching using a command-line interface plug-in
   2. BAutomated penetration testing using an AI agent
   3. CTranslation and summarization using a chatbot
   4. DFraud detection using transaction pattern analysis

   Show the answer

   Correct answer: C. Translation and summarization using a chatbot

   A chatbot can translate multilingual content and summarize the key security-relevant details for rapid incident communication.

8. Q8AI Governance, Risk, and Compliance

   Question 8. A software company adds an artificial intelligence (AI) feature that drafts quarterly performance feedback for managers. Employees later report they did not know AI was involved or what information the system used to generate comments. Which of the following BEST supports transparency?

   1. ARestrict the feature to senior managers so fewer employees are affected by the output.
   2. BStore generated feedback for a longer period so managers can compare it across review cycles.
   3. CTune the model until it produces more positive wording in the generated feedback.
   4. DProvide employees and managers with clear notice that AI is used, describe the main data sources, and state the system’s limitations.

   Show the answer

   Correct answer: D. Provide employees and managers with clear notice that AI is used, describe the main data sources, and state the system’s limitations.

   Transparency requires clear disclosure about AI use, the information involved, and limitations that may affect interpretation.

9. Q9Basic AI Concepts Related to Cybersecurity

   Question 9. A healthcare security trainer is testing a large language model (LLM) during a tabletop exercise. The trainer asks: "Explain how a help desk employee should respond to a suspected credential phishing call." The trainer provides no sample answers or demonstrations. Which of the following prompt engineering techniques is being used?

   1. AMulti-shot prompting
   2. BOne-shot prompting
   3. CPrompt templating
   4. DZero-shot prompting

   Show the answer

   Correct answer: D. Zero-shot prompting

   Zero-shot prompting asks the model to complete a task without providing examples in the prompt.

10. Q10Securing AI Systems

    Question 10. A claims-processing outsourcer uses an artificial intelligence (AI) service to summarize adjuster notes. The gateway requirements state that each request may include no more than 10 note files and no more than 25 MB of total uploaded data. Which of the following gateway controls BEST satisfies these requirements?

    1. APrompt firewall
    2. BInput quotas
    3. CRate limits
    4. DEndpoint access controls

    Show the answer

    Correct answer: B. Input quotas

    Input quotas can restrict both the quantity of submitted items and the total data size allowed per request.

11. Q11AI-assisted Security

    Question 11. A retail pharmacy chain discovers several social media posts claiming the company is closing all stores due to a patient data breach. The posts were created by a threat actor who knew the claim was false and wanted customers to visit a fake refund portal. Which of the following BEST classifies the false claim in the posts?

    1. ADisinformation
    2. BImpersonation
    3. CMisinformation
    4. DReconnaissance

    Show the answer

    Correct answer: A. Disinformation

    Disinformation is false information deliberately created or spread to deceive. The threat actor knowingly published a false breach claim to drive victims to a fake portal.

12. Q12AI Governance, Risk, and Compliance

    Question 12. A security analyst at a retailer uses an unsanctioned public artificial intelligence (AI) chatbot to summarize incident tickets. The tickets include customer email addresses, order numbers, and several authentication tokens copied from logs. Which of the following risks is MOST immediate?

    1. AAccidental data leakage through shadow AI
    2. BAutonomous systems failure
    3. CModel accuracy and performance degradation
    4. DIntroduction of bias in ticket classification

    Show the answer

    Correct answer: A. Accidental data leakage through shadow AI

    The analyst is using an unsanctioned AI service with sensitive operational and customer data, creating an immediate leakage risk.