Here are some approaches for red team attempts

1. Come up with input that potentially cause the LLM app to fail.
2. Pass this input to the LLM app and collect its output.
3. Check if the attempt was successful.

LIMITATIONS:-

• Inputs are explicitly decided upon and limited in number.
• Rule-based detection will decrease flexibility when checking if output matches expectations.

DEMO LLM APPLICATION:-

TESTING LINK : https://s172-29-78-176p8888.lab-aws-production.deeplearning.ai/notebooks/L4/L4_Red_teaming_LLMs_with_LLMs.ipynb