● AI Model Jailbreak Check

Is your model's API safe from prompt attacks?

Paste your model's HTTP endpoint. We run real generational jailbreak attacks against it — mutating and re-trying prompts across several generations — and tell you honestly whether any got through, with the exact prompts that worked.

Advanced: my API uses different field names

By default we POST {"prompt": "..."} and read the reply from a response field. If your API is shaped differently, adjust below.

I own this endpoint, or I'm authorized to test it.