CompTIA CY0-001: CompTIA SecAI+

Last checked October 22, 2026 against the official CompTIA exam page.

1. Question 1Securing AI systems

   An attacker submits a crafted prompt to a customer-facing assistant that tells it to ignore its instructions and reveal the system prompt. Which attack category is this?

   1. ASQL injection against the vector store
   2. BPrompt injection
   3. CDenial of service via packet flooding
   4. DCross-site scripting in the browser only
   Show answer and explanationHide answer

   Correct answer: B

   Prompt injection manipulates the model through input text to override instructions or leak hidden context. The other attacks target databases, browsers or network capacity rather than the model's instructions.

2. Question 2AI-assisted security

   A security operations centre receives thousands of alerts a day. The team wants an AI tool to group related alerts and rank them by likelihood of being a real incident. Which capability best fits?

   1. ASpeech recognition of recorded phone calls for transcript indexing
   2. BAlert correlation and risk scoring to prioritize analyst review
   3. CImage classification of scanned network diagrams to label device types
   4. DGenerating marketing copy for the SOC's public website and customer updates
   Show answer and explanationHide answer

   Correct answer: B

   Correlating related alerts and scoring their risk reduces noise and helps analysts focus on the likely incidents. The other capabilities are unrelated to SOC triage.

3. Question 3AI governance, risk, and compliance

   A company deploys an AI system that makes decisions affecting customers' access to a service. Which governance activity should happen before deployment?

   1. AA documented risk assessment covering impact, bias testing and human oversight
   2. BPublishing the model weights to attract users on a public repository before internal testing
   3. CSkipping review because the vendor is trusted and its SOC 2 report is on file
   4. DWaiting until a regulator asks for documentation, and then compiling the risk records later
   Show answer and explanationHide answer

   Correct answer: A

   A risk assessment that covers impact on customers, bias testing and human oversight is the standard pre-deployment governance step. Publishing weights, skipping review and waiting for a regulator leave the risks unmanaged.

4. Question 4Basic AI concepts related to cybersecurity

   A security analyst is asked to explain the difference between supervised and unsupervised learning for a fraud detection project. Which statement is correct?

   1. ASupervised learning trains on labeled examples, while unsupervised learning finds structure in unlabeled data
   2. BUnsupervised learning requires analysts to hand-label each record in the training set first
   3. CSupervised learning cannot be applied to classification tasks, since labeled outputs are continuous values rather than classes
   4. DUnsupervised learning guarantees the correct fraud label for each transaction once clusters form
   Show answer and explanationHide answer

   Correct answer: A

   Supervised learning uses labeled examples such as known fraud and legitimate transactions, while unsupervised methods such as clustering look for patterns without labels. The distractor about unsupervised labeling reverses the definition, and supervised learning is widely used for classification.

5. Question 5Securing AI systems

   A retrieval system pulls internal documents into prompts. A malicious contributor plants a document that instructs the model to send confidential data to an external address. Which risk does this describe?

   1. ACredential stuffing against the login page with breach data
   2. BModel extraction through repeated API calls
   3. CAdversarial patch attack on image classifiers
   4. DIndirect prompt injection through retrieved content
   Show answer and explanationHide answer

   Correct answer: D

   Indirect prompt injection hides instructions in content the model retrieves, so the model follows them without a direct attacker prompt. Model extraction, image patches and credential stuffing are different attack types.