This index does not restate the AI Act. It provides an interpretive reading of high-risk domains in order to identify where meaning must be bounded, justified, escalated, or refused.
Operational definition
The interpretive AI Act index is a navigation layer that groups sectoral maps according to domains where generative synthesis can become actionable, contestable, or institutionally sensitive. Its purpose is operational: to connect high-risk contexts to the constraints they require.
Why an index is necessary
Regulation names categories of risk, but governability requires a second layer: the structure of meaning that an AI system may overextend, simplify, or silently universalize. The index translates sectoral exposure into interpretive constraints, without confusing legal qualification with editorial architecture.
Domains covered by the index
- Employment and HR: criteria, exclusions, bias, and traceability.
- Education: thresholds, evidence, and legitimate non-action.
- Credit: factors, negations, justification, and temporality.
- Health: prudence, source hierarchy, limits, and human escalation.
- Legal and public sector: jurisdictions, exceptions, validity, transparency, and recourse.
- Biometrics and identity: identification, verification, surveillance, and prohibitions.
How to use the index
- Start from the sector where an output can become actionable rather than merely descriptive.
- Move from the sectoral page to the corresponding map of constraints.
- Relate each sectoral map back to phenomena, doctrine, and definitions.
- Use the index as a routing layer for governance, not as a substitute for domain expertise.
- Keep the distinction between regulatory compliance and interpretive stability explicit.
What this index prevents
- Treating all high-risk contexts as if they required the same warnings and the same editorial pattern.
- Reducing governance to a compliance label without operational constraints.
- Confusing sector naming with interpretive control.
- Leaving domain-sensitive outputs without a canonical route toward the relevant map.
Recommended links
Cross-domain governance core
Whatever the high-risk domain, the same core returns:
- an explicit canon for what prevails;
- response conditions that make prudence, non-response, or escalation possible;
- a proof of fidelity between source, synthesis, and perimeter;
- an interpretation trace that makes the decision contestable;
- enough observability to see whether the right surfaces are actually being activated.
This is why the index must remain tied to Q-Layer, Proof of fidelity, Interpretation trace, Interpretive observability, and Interpretive auditability of AI systems.
What this index should trigger
The index is not an end point. It should trigger more precise reading:
- by vertical when risk depends on a sector;
- by mechanism when the problem comes from smoothing, extrapolation, or authority confusion;
- by surface when the canon, governance files, versions, or error registries must be strengthened.
For the upstream layer, see Machine-first is not enough: why governance files change the reading regime, Site role, and Observations.