AISI
AI Safety Institute. The atlas uses this label for public institute-style bodies focused on AI safety testing, evaluation, assurance, or coordination.
Frontier Ai SafetyEvaluationAi Assurance
Glossary
Atlas terms and editorial flags
Short definitions for the labels that appear across entity cards, country pages, source records, and relationship maps.
Confidence level
A conservative editorial estimate of how strongly the current public source set supports a record or relationship.
Needs primary source
A flag for discovery records currently backed by research briefs or synthesis sources that should be upgraded with official pages, filings, standards, papers, or event records.
Dialogue channel
A convening, event series, working group, or organization that may support cross-border AI safety and governance discussion.
International DialogueDialogue Channel
Public safe to show
A Boolean marker used on relationships and evidence notes to avoid publishing sensitive, speculative, or insufficiently sourced claims.