CHECK POINT™

AI Guardrails Powered by Lakera
Knowledge Base
How a message flows through Lakera Guard
Custom System Prompt — CP0 Lakera Scan Before Activation
Settings — API keys & LLM provider Configured here; persist with “Save to .env”
Grade attacks with a stronger model so a weak target isn't judging its own output. Used only by the one-shot judge.
Custom RAG Docs — Scanned by Lakera on Retrieval PDF · DOCX · TXT · MD · CSV · JSON · max 2 MB / file · max 10 files
Upload › › Store › CP2 · scan on retrieval › LLM
No files uploaded yet.
Hi! I'm ShopEase's AI customer service assistant. How can I help you today?
Security Trace
Guard pipeline
Input › › › LLM › › Reply
CP0 — System Prompt Scan Default
One-time scan on prompt upload
CP1 — User Input Pending
Checks prompt injection & jailbreaks
CP2 — RAG Documents Pending
Checks retrieved knowledge-base docs
CP3 — LLM Output Pending
Checks AI response before delivery