AI Trust Glossary · Canonical Definition
BorealisMark Scoring
BTS Credit Rating
A letter grade derived from an AI agent's BTS trust score, ranging from AAA+ (near-perfect trust, score 980+) to FLAGGED (high-risk agent, score below 500). The human-readable companion to the numeric BM Score - one glance answers whether an agent should be trusted.
Complete rating scale
| Rating | Score threshold | Interpretation |
|---|---|---|
| AAA+ | 980 - 1000 | Near-perfect behavioral trust |
| AAA | 950 - 979 | Excellent - production-grade |
| AA+ | 920 - 949 | Very high trust |
| AA | 880 - 919 | High trust |
| A+ | 840 - 879 | Strong trust |
| A | 800 - 839 | Good trust |
| BBB+ | 750 - 799 | Adequate - monitor closely |
| BBB | 700 - 749 | Marginal - use with oversight |
| UNRATED | 500 - 699 | Insufficient data or poor performance |
| FLAGGED | Below 500 | High-risk - do not deploy |
Why the financial credit rating model
Bond credit ratings (S&P's AAA through D) are universally understood by procurement, legal, and executive audiences. When a compliance officer asks "is this AI agent safe to deploy in a regulated environment?" the answer "AAA" communicates the same confidence as it does in a bond offering memorandum. Numeric scores (96.1/100) require context. Letter grades communicate trust tiers instantly across disciplines - the target audience for AI trust certification is not data scientists, it is the people who sign deployment approvals.
Tier-based scoring ceilings
The maximum achievable credit rating varies by license tier. Free tier BTS keys have a trust ceiling of BBB+ (max BM Score 65 displayed, raw 650/1000). Pro tier keys have a ceiling of A (max BM Score 85, raw 850/1000). Sidecar-verified agents are uncapped - they can achieve AAA+. The ceiling exists because self-reported telemetry is unverified. Sidecar-verified telemetry carries an independent observer's signature, justifying the uncapped score.
See also