> ## Documentation Index
> Fetch the complete documentation index at: https://project-brain.rocheaiops.net/llms.txt
> Use this file to discover all available pages before exploring further.

# Trust and authority

> Authority and confidence are separate, explicit dimensions of an interpretation.

Authority identifies the source’s right to establish a fact for the relevant context. Confidence captures uncertainty in an interpretation, extraction, or claim. An owner can authoritatively state uncertainty; a technically plausible document can still have lower authority for a personal preference.

| Source                           | Automatic canonical trust?        | Typical treatment           |
| -------------------------------- | --------------------------------- | --------------------------- |
| Explicit owner teaching          | No bypass; validated structurally | May become active           |
| Owner statement with uncertainty | No                                | Preserve uncertainty        |
| Document source                  | No                                | Knowledge/citation evidence |
| Model inference                  | No                                | Candidate only              |
| Gemini pretrained knowledge      | No                                | Provider context only       |
| Future web/sensor input          | No                                | Future evidence pipeline    |

No numeric authority weights are documented because none are implemented. See [the cognitive model](/architecture/cognitive-model).