Service

Adversarial Fuzzer

FULL PLAN ONLY

Real generational prompt-mutation red-teaming against an HTTP model endpoint you supply. Targets are restricted to public addresses (no probing internal infrastructure); scoring is a documented refusal-phrase heuristic, not a verified jailbreak oracle.

No per-call charge -- unlimited under your subscription

Worked example

Start Fuzz

curl -X POST https://cainstudio.online/adversarial-fuzzer/fuzz \
  -H "X-API-Key: $CAIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model_id": "...", "target_url": "...", "prompt_field": "prompt", "response_text_field": "response", "seed_prompts": [], "generations": 5}'

Try it live in your dashboard → See pricing →