Methodology

Trust methodology

Clairist computes trust at two layers: a live per-system Trust State and a workspace trust summary. This page describes each layer as implemented.

1. Live per-system Trust State

On each AI system page, Clairist computes a live score from system-level inputs. This is separate from workspace analytics and is used for system-specific trust decisions.

  • 20 ptsSystem definition Name, description, provider, and model fields.
  • 30 ptsEvidence coverage Evidence items attached to that system.
  • 40 ptsDisclosure coverage Required disclosures completed, with evidence linkage checks.
  • 10 ptsFreshness Recency of system activity.

2. Workspace trust summary

Trust analytics and the Overview workspace trust headline use workspace analytics from the trust cache pipeline—not the live per-system engine.

  • Source Trust summary values are read from trust_insights_cache.
  • Freshness Analytics freshness is tracked from cache timestamps and surfaced as current, degraded, or unavailable.
  • Legitimacy fields Verification tier and evidence freshness are resolved from team trust settings and team health summary data.

How these layers relate

These layers are intentionally distinct:

  • The AI system page Trust State is the live system-level evaluator.
  • The workspace trust summary is an aggregated analytics headline.
  • A system can have a live Trust State before workspace analytics refresh, so temporary differences can occur.

How to use this in audits

For defensible reviews, use both surfaces together:

  • System decisions Use live per-system Trust State factors and actions.
  • Portfolio posture Use workspace trust summary and trend and insight surfaces.
  • Verification context Include verification tier and freshness states when interpreting workspace scores.

Interpretation

Trust State, workspace trust summaries, and methodology descriptions are product documentation only. They are not legal or compliance advice, not a certification, and not a guarantee that any AI system is safe or lawful. Beta customers should treat this page as beta-ready draft documentation to be reviewed alongside counsel where material decisions depend on it.

Trust CenterBack to home — trust signals