Hack OpenAI E-Banking ChatBot: A Hands On Lakera Guard Experience

[Shay_Levin]

Learn how Lakera Guard protects AI applications from prompt injection attacks, data leakage, and SQL injection

To view this video, you must Accept and consent to the use of cookies.

Accept

 🚀 TRY THE DEMO YOURSELF:

1. Visit CheckMates Labs: https://community.checkpoint.com/t5/CheckMates-Labs/ct-p/checkmates-labs?from=community_slider

2. Find the "Lakera Guard Demo" lab

3. Launch the interactive environment

4. Follow the step-by-step guide: https://checkmatesstorage.blob.core.windows.net/public-json/DEMO_GUIDE.html

🎯 What You'll See:

• Direct prompt injection attacks and detection

• Stored/indirect prompt injection via database fields

• MCP (Model Context Protocol) poisoned document attacks

• SQL injection detection

• Real-time bidirectional scanning (input + output)

• Alert mode vs Blocking mode comparison

📖 The Story: Follow "Sara Mitchell" as she attempts to manipulate an AI banking assistant to:

1. Expose her hidden risk score and internal notes

2. Modify her customer profile through a poisoned document

3. Fraudulently obtain a $450,000 loan

🛡️ Lakera Guard Features Demonstrated: ✅ Prompt injection detection ✅ Jailbreak attempt detection ✅ Data leakage prevention ✅ PII exposure detection ✅ SQL injection detection ✅ Content moderation

[sjni01]

@Shay_Levin Awesome, regards from South America Master!