AI Tools: Promptfoo Red-Team Test Suites

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-tools-ai-promptfoo-redteam-r10a4-creators

1. What is the primary function of Promptfoo's red-team plugins?

   1. To automatically deploy AI applications to production servers
   2. To generate creative marketing copy for AI products
   3. To probe your app with adversarial prompts and grade responses against safety policies
   4. To translate prompts into multiple languages for international users

2. Which action is recommended when a high-severity red-team test shows regression compared to the last release?

   1. Increase the model's temperature setting to improve creativity
   2. Block the deployment until the vulnerability is addressed
   3. Downgrade to the previous test suite version
   4. Deploy with a warning notification to users

3. Why is it important to refresh red-team prompts quarterly?

   1. Because quarterly is required by most security certifications
   2. Because AI models become more expensive over time
   3. To prevent attackers from outpacing your test suite
   4. To ensure tests run faster on older hardware

4. Which of the following is a capability that AI provides in red-team testing?

   1. Covering every possible security threat
   2. Enabling jailbreak, PII, and harmful content plugins
   3. Replacing human red teamers entirely
   4. Fixing ambiguity in safety policies automatically

5. What does it mean to 'tie the suite to CI' in the context of red-team testing?

   1. To store test results in a separate database from other CI metrics
   2. To integrate red-team tests into the continuous integration pipeline so they run automatically with each build
   3. To require two separate approvals before running any test
   4. To run tests only when developers manually trigger them

6. What is a 'jailbreak' in the context of AI safety testing?

   1. An open-source alternative to closed-source AI APIs
   2. A method to make AI models run faster on mobile devices
   3. A tool for compressing large language models
   4. A prompt designed to bypass an AI's safety guidelines and elicit restricted outputs

7. Which of the following is explicitly listed as something AI CANNOT do in red-team testing?

   1. Enable jailbreak and harmful content plugins
   2. Integrate with CI/CD pipelines
   3. Replace human red teamers
   4. Track regression over software releases

8. What is the purpose of tracking regression over releases in red-team testing?

   1. To identify when new code changes introduce security vulnerabilities that weren't present before
   2. To calculate the total number of prompts processed
   3. To measure how much money the testing infrastructure costs
   4. To determine which developer wrote the most buggy code

9. What type of information does PII stand for that red-team plugins test for?

   1. Program Integration Index
   2. Personally Identifiable Information
   3. Prompt Iteration Interface
   4. Public Intelligence Index

10. What is a fundamental limitation of automated red-team testing?

    1. It cannot cover every possible threat
    2. It cannot generate enough test prompts
    3. It cannot be integrated with modern development workflows
    4. It cannot distinguish between harmful and harmless content

11. Why should organizations maintain human red teamers despite using automated tools like Promptfoo?

    1. Automation has made human oversight completely unnecessary
    2. AI tools cannot replace human creativity and judgment in discovering novel attack vectors
    3. Human red teamers are less expensive than automated tools
    4. Humans can type faster than AI systems

12. What happens when safety policies are ambiguous in the context of AI deployment?

    1. AI cannot fix policy ambiguity—this requires human clarification
    2. AI automatically resolves the ambiguity by choosing the safest option
    3. The red-team plugin flags it as a pass regardless of response
    4. The system defaults to allowing all content

13. What is the relationship between red-team testing and CI/CD pipelines?

    1. Red-team tests should only run on development machines, never in production pipelines
    2. CI pipelines cannot handle the computational load of red-team testing
    3. Red-team tests replace the need for any other security testing in CI
    4. Red-team tests can be integrated into CI to automatically run with each code change and block insecure deployments

14. If an AI system passes all red-team tests, does this guarantee it is completely safe?

    1. Yes, if all tests pass, the system is 100% secure
    2. Yes, but only if the tests were run in production
    3. No, red-team tests cannot cover every possible threat
    4. No, but the tests should not be run again for six months

15. What is the primary goal of grading responses against safety policies in red-team testing?

    1. To train the AI to generate more creative responses
    2. To measure how quickly the model responds to prompts
    3. To determine whether the AI's outputs comply with defined safety guidelines
    4. To optimize the cost of running inference