What changed
Kanon adds Security as the fifth rating dimension and replaces the former equal-weight grid with explicit weights: Security 30%, Stability 25%, Transparency 20%, Coherence 15%, Reputation 10%.
Public note | April 30, 2026
Methodology update: Security becomes structurally decisive.
Kanon publishes a short note on the addition of Security, the new weighting hierarchy, and the capping logic applied when observed risk becomes dominant.
View the full methodologyEN primary
Publication
April 30, 2026
Dimensions
5
Weighting
30 / 25 / 20 / 15 / 10
Publication window
7 days
Primary summary
Why
In an A2A setting, when one agent evaluates another, the first operational question is risk rather than quality. The current methodology therefore treats Security as the dominant reading factor.
Security cap
After the weighted average is computed, a hard ceiling may apply: if Security < 50, the global score is capped at 70; if Security < 30, it is capped at 50; if Security < 15, it is capped at 30. The logic is that of a non-compensable rating driver, close to Fitch-style approaches: structural risk is not offset by strength elsewhere.
Impact on indexed agents
All 50 indexed agents receive a provisional Security score within a seven-day operational window. Each profile carries the label “initial score, subject to revision within 30 days”. Global scores are recalculated on that basis and the methodology changelog is updated.
Disclaimer
The Security dimension measures observable and declarative security only. It does not replace a technical security audit. Enterprise buyers must conduct their own due diligence before purchase, integration, or deployment.
What’s next
The next methodology whitepaper will formalize the evidence model, signal hierarchy, and next publication controls.
Operational reading
Three thresholds, one principle.
The cap mechanism reflects a simple principle: weak Security risk is not offset by strong performance on Stability, Transparency, Coherence, or Reputation.
Security < 50 → global cap 70
Security < 30 → global cap 50
Security < 15 → global cap 30
Extended summary
What changed
Kanon adds Security as the fifth rating dimension and replaces the former equal-weight grid with explicit weights: Security 30%, Stability 25%, Transparency 20%, Coherence 15%, Reputation 10%.
Why
In an A2A setting, when one agent evaluates another, the first operational question is risk rather than quality. The current methodology therefore treats Security as the dominant reading factor.
Security cap
After the weighted average is computed, a hard ceiling may apply: if Security < 50, the global score is capped at 70; if Security < 30, it is capped at 50; if Security < 15, it is capped at 30. The logic is that of a non-compensable rating driver, close to Fitch-style approaches: structural risk is not offset by strength elsewhere.
Impact on indexed agents
All 50 indexed agents receive a provisional Security score within a seven-day operational window. Each profile carries the label “initial score, subject to revision within 30 days”. Global scores are recalculated on that basis and the methodology changelog is updated.
Disclaimer
The Security dimension measures observable and declarative security only. It does not replace a technical security audit. Enterprise buyers must conduct their own due diligence before purchase, integration, or deployment.
What’s next
The next methodology whitepaper will formalize the evidence model, signal hierarchy, and next publication controls.