AI System Fact Sheet

Trooth's AI powers two product features: Kyrie, a conversational assistant, and questionnaire drafting, which proposes answers to security and vendor questionnaires. Both are assistive: they draft and explain, and a person decides what to do with the result.

These features are built on third-party large language models accessed through a managed gateway, not models Trooth trains itself. We select models from established providers (such as OpenAI and Anthropic) for reasoning quality and safety, and we may change the underlying model as better options become available. Model routing is a Trooth implementation detail; the human oversight and data commitments on this page hold regardless of which model is in use.

To answer a question or draft a response, a model sees only the context needed for that task: your prompt, relevant evidence and posture data Trooth has already witnessed, and framework reference material. It does not receive your credentials or raw integration secrets.

Trooth does not use your data to train AI models, and prompts are not contributed to provider training. See our AI Use Policy for the full stance.

• Output reflects the evidence and context available at the time and can be out of date.
• The models reason over language, not a ground-truth audit; they can miss nuance in your environment.
• Drafts are a starting point for security and compliance work, not a substitute for professional judgment.

Like all large language models, the systems behind Kyrie and questionnaire drafting can produce output that sounds confident but is inaccurate or fabricated. Output may be inaccurate and should be reviewed by a person before you rely on or submit it. Trooth labels AI-generated content in the product and keeps a human in control of anything that has real effect.