Live security demo

Protect LLM chat with Lakera Guard

A public demo showing how to gate OpenAI requests with Lakera Guard prompt injection detection. Try safe prompts and attack examples side by side.

Open interactive demoLakera Guard docs

How it works

  1. Enter a prompt or pick a sample test case
  2. Lakera Guard screens the input for prompt injection
  3. Safe prompts get an OpenAI response; flagged prompts are blocked
User → Lakera Guard → if safe → OpenAI → Response
                    → if flagged → Block + show details

Input screening

Every prompt is sent to Lakera Guard before it reaches OpenAI, catching injection attempts early.

Fail-closed blocking

Flagged prompts are blocked server-side. OpenAI is never called when a threat is detected.

Transparent results

See safe vs. blocked status, detector breakdown, and whether an AI response was generated.

Ready to test?

Use built-in sample prompts including normal requests, direct injection, and indirect RAG-style attacks.

Try the demo