AAgentProof

AI Landscape Radar · how it works

Trust boundaries

How often each source category is reviewed, and what stays fixed once a report is generated.

Source categories have a documented target cadence

Each category of public source AgentProof watches has a documented target review cadence — most monthly, regulation and governance quarterly — with emergency reviews when something material lands. Reviews are deliberate, on-demand and human-led, not a live crawl.

Old reports stay frozen

A report is pinned to the methodology version it was scored under. A later review never reaches back and silently re-scores work you already received — the original report stays exactly as issued.

Reassessment may be recommended, never forced

When a review leads to a change that could affect an existing report, you see a clear 'Reassessment recommended' notice. You decide whether and when to re-run — nothing happens automatically.

No silent rescoring

Methodology changes are versioned and visible. The version your report was scored under is always shown, so you can always tell what changed and when.

← Back to AI Radar summary