Professional AI Red Team Reporting COASP Practice Questions (Page 9)

Questions 41–45

1. 41foundation · easy

   Why is it important to document the attack path in a technical finding?

   • ATo allow the remediation team to reproduce the vulnerability and verify the fix
   • BTo provide a detailed log of every command executed during the assessment
   • CTo prove that the red team was authorized to perform the test
   • DTo assign blame to the developer who wrote the vulnerable code
   Reveal Solution Discussion Select an answer first
2. 42foundation · easy

   Why should recommendations in an AI red team report be prioritized?

   • ATo ensure that all recommendations are implemented at the same time
   • BTo help the organization focus on the most critical risks first
   • CTo satisfy compliance requirements that mandate a specific order
   • DTo make the report longer and more detailed
   Reveal Solution Discussion Select an answer first
4. 43expert · hard

   A red team report for a financial trading firm's AI-based market prediction model identified a data poisoning vulnerability that could cause the model to make incorrect trades. The firm has a very limited budget and cannot afford to retrain the model frequently. The report must provide recommendations that are both effective and cost-conscious. What is the best recommendation?

   • AImplement a data provenance and validation pipeline to ensure that training data is trustworthy, and add a human-in-the-loop review for high-value trades.
   • BRetrain the model daily with fresh market data to reduce the impact of any poisoned data.
   • CDisable the AI-based trading system and revert to manual trading decisions.
   • DIncrease the model's complexity to make it more difficult for attackers to poison the training data.
   Reveal Solution Discussion Select an answer first
5. 44application · medium

   During a red team assessment, you successfully extracted a small set of personally identifiable information (PII) from a language model using a prompt injection attack. You need to include this evidence in the report. What is the most appropriate way to present the evidence?

   • AShow the exact prompt used and a redacted version of the model's output, with the PII partially masked.
   • BProvide the evidence only verbally to the client's security team and do not include it in the written report.
   • CDescribe the attack in the methodology section but do not include any output, to avoid any risk of data exposure.
   • DInclude the full unredacted output in the report but mark the document as confidential.
   Reveal Solution Discussion Select an answer first
6. 45application · medium

   A red team report for a retail company's AI chatbot identified a prompt injection vulnerability that could be used to extract customer PII. The company has limited resources and wants to address the most critical issues first. Which recommendation should the report prioritize?

   • ADisable the chatbot's access to PII entirely and require all customer data requests to go through human agents.
   • BRetrain the chatbot model on a larger dataset to improve its general understanding and reduce its susceptibility to unusual prompts.
   • CConduct a full penetration test of the entire IT infrastructure to identify other vulnerabilities that may also lead to PII exposure.
   • DImplement input sanitization and output filtering on the chatbot to block prompt injection attempts, and restrict the chatbot's access to PII.
   Reveal Solution Discussion Select an answer first