What it is
A focused AI security publication and public repository posture index with methodology, freshness, evidence, and limitations attached.
About
The public evidence layer for AI security posture
AISecurityBase combines current reporting, static repository rankings, and selected case files so practitioners can test security claims against visible evidence.
The mission
Make useful security evidence legible before claims outrun controls.
AI security moves across code, models, infrastructure, policy, and real-world incidents. This publication connects those surfaces without blurring reporting, assessment, and commercial work.
What it is not
A certification program, a pay-to-rank directory, or a generic feed of rewritten announcements.
Product boundary
thethermark is the separate commercial assessment product. AISecurityBase owns public presentation, editorial review, corrections, publishing, and sponsorship surfaces.
How trust is earned01 02 03 04
The evidence chain stays visible
Static first, evidence attached
A versioned static assessment is enough for inclusion when readers can inspect its scope, dimensions, freshness, and limitations.
Assurance levels stay distinct
Static Assessed, Deep Reviewed, and Runtime Assessed labels describe different evidence levels. Runtime evidence never silently changes the static score.
Corrections are free
Maintainers can submit evidence, corrections, and responses without payment or a commercial relationship.
Sponsorship never buys rank
Sponsored placements sit outside ranking order, findings, assurance badges, corrections, and editorial conclusions.