Chaos engineering for AI APIs
Automatically uncover adversarial failures before they reach production.
e.g. https://api.example.com/v1/chat or http://localhost:8000/mock-vulnerable-api
Optional — used to tailor adversarial prompts when using Gradient AI.
For OpenAI/Gradient endpoints when using messages format.
For real API demos (e.g. Gradient). Not stored; sent only with the scan request.