Probes front-end chat and customer-facing AI. Sends natural language inputs and scores responses.
Probes back-end scoring APIs. Sends matched pairs varying protected characteristics to detect differential outcomes.
Conversational mode — best for customer service bots, virtual assistants, and triage tools with a chat interface.
Decision model mode — best for credit scoring, candidate ranking, and triage classification engines with a REST API.
The demo uses pre-configured endpoints. Switch to Live to connect your own AI tool.
Enter your AI tool's real endpoint details. Credentials are sent server-side and never exposed in the browser.
Request / response mapping
Input field mapping
Map your API's input fields to the protected characteristics TrustHuman will vary in matched-pair testing. TrustHuman sends two identical requests differing only in the mapped characteristic and compares the outputs.
Output field mapping