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. Letter bands A through F. Reproducible to the byte.
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, reproducible to the byte, and presented in letter bands from A through F.
No institution reaches the A-band. Credit ratings put 49 there.
of 321 paired
institutions
Credit ratings put
49 here.
Across the 321 institutions in the substrate that hold at least one major credit rating, the two instruments measure different things. Credit ratings cluster at A and B (244 institutions, 76 percent). MAR clusters at C (184, 57 percent). The instruments tell different stories about the same institutions because they answer different questions, and the shape of the divergence is what governance risk looks like when measured directly.
Only 1 in 3 triple-rated institutions get the same notch from all three agencies.
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.
Banks strongest. Pension funds weakest.
Eight sectors. Each row normalised to that sector's rated institutions. Hover or tap any segment for the exact count. Banks lead with 26 percent in B-band and only 2 percent in E-band. Pension funds are the inverse: 2 percent in B-band, 17 percent in E or F.
Europe leads. Measurably.
Europe shows the cleanest distribution: 22 percent in B-band, 1 institution in E-band, none in F. The EU regulatory regime, with DORA in force and the EU AI Act phasing in, correlates with measurably better governance than any other region. Middle East and Africa shows the weakest distribution, with 2 institutions in F-band deep failure.
across agencies
observable governance
halt mechanism
dominant vendor
identity unobservable
SHA256 e5250de8e9de07d6
See where you score, who scores like you, and where the gap is. The diagnostic engagement converts the score into a remediation programme. Continuous Intelligence keeps the picture current as the substrate refreshes.
Independent cross-jurisdiction benchmark on AI governance, scored monthly, sealed for reproducibility, evidence base traceable to source. Supervisor toolkits arriving 2026.
Governance risk that credit ratings cannot see. AI agent concentration, halt-mechanism gaps, vendor-mesh contagion. Material to enterprise valuation, and currently unpriced.
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.
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].
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