In this exercise, you will engage with gen AI (or public discussion forums as well) to craft prompts designed to challenge and evaluate the resilience of our newly developed EU AI Act policy.
The goal is to explore the boundaries of the policy across all configured areas, identifying potential gaps, edge cases, or unintended behaviours.
Once gaps are identified, we will work together to refine and fine-tune the policy to ensure its effectiveness and resilience in real-world AI applications.
After this exercise, the updated policy will be exported and shared with all paladins!
You are encouraged to think creatively and critically, simulating real-world scenarios where AI outputs might conflict with regulatory guidelines. Your task is not only to generate prompts but also to observe and document how the AI responds, highlighting areas where the policy performs as intended and where it may require adjustments.
NOTE: You can test the Homegrown AI App policy (Prompt Injection) using Playground directly from the console!