GOVERNANCE · VERSION 1.0

Challengeable by design.

AI Risk Trajectory publishes its evidence and correction rules because an authoritative-looking interface is not a substitute for governance. External expert review has not yet been completed, and the current classifications are not presented as scientific consensus.

Source policy

News and secondary coverage may identify candidate evidence, but material classifications should prefer primary or authoritative underlying sources.

• Primary research and evaluations
• Independent evaluators
• Government and international scientific assessments
• Standards and risk-management bodies
• Provider system cards and threat intelligence when clearly labelled

Review policy

Current public snapshot is a research preview and has not undergone formal external peer review.

Material state changes should receive primary analyst review and a second reviewer; high-consequence or contested changes should also receive relevant domain-specialist review.

Reviewer disagreement should be recorded as disagreement, not averaged into a synthetic score.

Corrections

Substantive corrections create a visible correction record. Historical snapshots are not silently rewritten; superseded states remain archived.

Anyone may challenge a classification, source, causal relationship, control assessment, methodology rule or missing evidence through the structured challenge form.

AI use

AI can assist discovery, extraction, explanation and consistency checking. It cannot autonomously publish material state changes or invent probabilities and reviewer agreement.

External review status

Not yet completed. The product should not describe its classifications as scientific consensus until qualified external review has been completed and published.

Challenges can be submitted from the Governance section inside the interactive tool.