Governance Benchmark Index
One score: a system-level read of whether an autonomous system's governance is adequately designed, on a scale where lower is stronger.
DSI 001 Version 1.0. Effective [25 June 2026].
A numeric scale, not a letter grade, because institutional users need granularity. Lower means stronger assessed governance posture. The Certified and Compliant tiers additionally require operational coherence and qualifying evidence; the GBI band is necessary, not sufficient.
What the GBI is
The Governance Benchmark Index is the scoring output of a DSI 001 assessment: a system-level read of whether a specific autonomous system's governance is adequately designed, expressed in a form institutional audiences can compare.
Boards, insurers, investors and regulators cannot each re-run a full governance assessment every time they meet an autonomous system. They need governance posture in a form that is comparable across systems and readable without reproducing the methodology. The GBI does that, on the same logic as an independent audit: an independently produced, comparably expressed signal that lets risk be assessed rather than avoided.
It assesses whether governance is adequately designed. It is not a rating, and it is not a recommendation to buy, sell, lend, insure or invest.
The six dimensions
Each system is read across six governance dimensions: structural properties of the decision supply chain that determine whether autonomous outcomes can be traced, attributed and governed.
Each dimension is scored on the same scale. The GBI aggregates them into a single comparable signal, alongside the full dimensional profile.
Composite and dimensional profile
An assessment produces two outputs, and both are required. The composite produces comparability: whether the system is above or below a threshold. The dimensional profile produces actionability: which governance investments will move the score and which risks are structural rather than operational.
The public methodology sets out the structure of the GBI: the six dimensions, the scale, the certification tiers, the multiplier categories, and what a valid result must contain.
Assessor implementation materials govern scoring procedures, calibration, quality control and consistency across accredited assessors. The index structure is public. The assessor materials preserve discipline in how the methodology is applied.
Why the GBI is not simply additive
Autonomous systems rarely fail through a single isolated weakness. The six dimensions are not independent: a weakness in one can compound a weakness in another, and the combined exposure is greater than either alone. Where compound weaknesses are present, multiplier logic adjusts the composite to reflect the amplified systemic exposure. A multiplied score is not a simple average. The primary multipliers describe governance conditions, not merely calculations.
High operational autonomy with inadequate liability architecture. The volume of autonomous-action consequences is high and the liability governance is inadequate to manage them.
Significant autonomy with inadequate contract infrastructure, including dependence on upstream providers whose termination rights or unilateral model changes could remove governance authority from the deploying organisation.
High commercial dependency with inadequate liability architecture, leaving the system structurally resistant to remediation.
High commercial dependency combined with inadequate contract infrastructure, where deployment relies on a single external AI provider and the deploying organisation's customers carry material government-adjacent concentration: entities holding government, defence or regulated-procurement obligations. A government action against the provider, such as a supply-chain designation, sanction or similar restriction, can convert that customer concentration into existential revenue loss, because those customers cannot continue using a designated provider's product, where the contract architecture does not address the cascade mechanism.
Designed, and exercised
The GBI evaluates design adequacy: whether governance architecture is correctly structured for the risk profile. The Governance Coherence Index evaluates operational coherence: whether that architecture was actually exercised across a defined window, the way SOC 2 Type II tests operating effectiveness.
The GBI answers whether governance was designed correctly. The GCI answers whether it operated correctly. Compliant and Certified determinations require both.
Evidence quality and confidence
The score communicates governance posture. The evidence quality behind it communicates the confidence warranted in that posture. Two identical scores can carry different institutional confidence depending on the tier of evidence supporting them.
The standard for evidence is not what the organisation says about its governance. It is what the governance infrastructure produces. The Evidence Infrastructure Standard defines the minimum structure of a qualifying Tier 1 record. Management representation is Tier 4 evidence: formal representation where no contemporaneous record exists. It is not sufficient evidence for a Compliant or Certified determination.
What a GBI result must include
A score on its own is not a DSI 001 result. Every GBI output is accompanied by the profile that makes it usable.
Use the GBI in an assessment
A GBI is issued only through a DSI 001 assessment by an accredited assessor, and it arrives as the full governance result described above, not a bare score. Self-scored or indicative figures are not DSI 001 results.