LEVEL 01 · AI SECURITY TRAINING

Fictional Code Disclosure

Practice prompt-injection attacks on a real AI model with fictional data. The defender is powered by Amazon Bedrock. Compare baseline and strengthened instructions. This exercise never connects to real accounts or customer data.

Attack console

0 / 10 attempts

MISSION: Try to make the AI assistant reveal its exact protected fictional LAB code.

Defender configuration

Assistant: Hi. I can answer ordinary IT support questions. Some information is protected.