Transparency
AI Disclosure
Last updated:
What is the AI persona?
The AI persona is a large-language-model agent that plays the role of a customer in scripted roleplay scenarios. It runs in two roles:
- Persona Agent — replies to your messages in real time during the conversation.
- Evaluator Agent — after the conversation ends, scores it against a configurable rubric and returns a structured report.
Where your data goes
| Step | Provider | What is sent |
|---|---|---|
| Roleplay (chat) | Anthropic (default) | Scenario brief, persona spec, your messages. |
| Pre-stream moderation | OpenAI | The text of the message about to be sent (no account context). |
| Scoring | Anthropic (default) or OpenAI | Full conversation transcript and rubric, no PII beyond your role and language. |
| Storage | Cloudflare R2 | Encrypted transcripts, KB documents, certification PDFs. |
What is not used for training
All LLM providers are bound by Data Processing Agreements that prohibit using customer data to train, fine-tune, or improve their base models. Your messages, scenarios, and scores stay inside the inference loop — they are not retained by the provider beyond the limits set in those agreements.
Limits and human oversight
- AI replies and scores can be wrong. Treat them as advisory.
- Certification requires a human approval click — the AI score does not by itself grant or revoke certification.
- A persistent disclosure banner appears at the top of every chat.
- You can withdraw recording consent at any time from your profile page.
- You can export or delete your data at any time from your profile page.