MAR500 · Meridian Autonomy Rating

The AI governance rating for regulated financial institutions. Issuer does not pay.

MAR500 measures how 543 regulated financial institutions in 66 countries govern their AI agents.

543
Regulated institutions in the sealed substrate
66
Countries covered
95,876
AI agents identified at the agent level

The Meridian Autonomy Rating (MAR) measures AI risk exposure in regulated financial institutions. Governance is the means; exposure is the end. MAR constructs governance topologies from public evidence using a predeclared scanning pipeline, and rates the exposure they produce. Credit ratings rate institutional default risk. MAR rates AI risk exposure at five levels: institutional, sectoral, geographic, vendor (cross-industry), and systemic. Credit ratings reach one of those levels. MAR reaches all five.

MAR500 applies the rating across 543 institutions in 66 countries on a sealed monthly substrate, scored along five components.

Notch spread among S&P, Moody's, Fitch

Only 1 in 3 triple-rated institutions get the same notch from all three agencies.

0 · unanimous 1 · close 2 · meaningful split 3 · large gap
0 · 67 1 · 89 2 · 46

Across the 206 institutions in the substrate rated by all three of the major credit agencies, only 67 receive identical notches from each. The remaining 139 carry disagreement: 89 within one notch, 46 spread two notches apart, four sitting across three notches. The mean spread is 0.94 with a standard deviation of 0.79. The instrument the market treats as the standard for credit risk reaches internal consensus on roughly one institution in three. MAR is deterministic by construction. There is one Meridian rating per institution because there is one Meridian.

543 institutions · 95,876 agents · 626,390 edges · 66 countries · sealed substrate v13.1.0
Six supporting findings
Intended audience
Institutions

Continuous Intelligence keeps the picture current as the substrate refreshes.

Supervisors

Independent cross-jurisdiction benchmark on AI governance, scored monthly, sealed for reproducibility, evidence base traceable to source. Supervisor toolkits arriving 2026.

Investors

Governance risk that credit ratings cannot see. AI agent concentration, halt-mechanism gaps, vendor-mesh contagion. Material to enterprise valuation, and currently unpriced.

Vendors

Independent view of how regulated counterparties depend on your products, the governance contribution those products carry, and where you sit relative to your peer set. Anonymised in public surfaces.

Section 6 · Computational Simulation

Hash-anchored hermetic simulation against the sealed substrate. Cascade analysis, algebraic connectivity, vendor cross-substrate persistence, and the Schrems II outlier model. Available to qualified institutional counterparties under non-disclosure agreement. Inquiries to [email protected].

MAR® Rating Methodology

The rating methodology that operates on this substrate will be documented in The Stationary Sea (Part 3: Rating Methodology), forthcoming. Specific calibrations are sealed via cryptographic manifest commitments. Manifest fingerprint · c545569ef34f35ba2b22821a99dcc0fd6c73a077863dbf6e0cde5946a18a85cf