PUBLIC REGISTRY
Sprinkling Act Public Registry
Three DOI-indexed research artefacts, the April 2026 anonymous cohort (50 entries), and client assessments, surfaced together with badge codes and DOIs. Each entry is independently resolvable.
Sprinkling Act publications
Indexed research · timestamped · openly licensed
Each publication is deposited on Zenodo with a permanent DOI, anchored on the Bitcoin blockchain via OpenTimestamps, and carries a verifiable badge resolvable from this registry. These are the validated artefacts that distinguish DOI/OTS entries from the other organisations listed below.
April 2026
EU AI Act Readiness Report — April 2026
What We Found Screening 50 European AI Companies
May 2026
Annex A — The Deployer Multiplier
Three downstream segments where the AI Act cascade becomes visible
May 2026
Discussion Paper · The AI Act as a Third Structural Pole
A Constrained Retrospective on the Predictive Frame (2015 to 2026)
April 2026 screening cohort
50 European AI companies · 4 sectors · 15+ countries
Aggregates sourced verbatim from the published April 2026 report. No company name disclosed, no individual score, no individual country. Privacy preserved by construction. The section below mirrors exactly what the DOI-indexed report already places in the public domain.
74%
Estimated high-risk
37/50
96%
No public AI Act position
48/50
44%
Art. 50 obligations unaddressed
22/50
0/14
Certified with AI Act mapped
100% gap
Sector breakdown
0 entries
Each entry is dated and carries the scope it covers. Re-scoring is performed by an analyst when a regulatory or system change is reported, not automatically.
INDEX UNDER CONSTRUCTION
What this index can say, and from when
The index aggregates what the surveys establish. Its value to a reader is a position: where do I stand. No public dataset allows that position to be calibrated. Eurostat counts adoption as yes or no, published experiments measure laboratory tasks, and the open datasets available cover entry into and exit from adoption. Nobody measures chains, hand-offs, rework time or named authority. The N cannot be borrowed anywhere. It has to be built.
So we state the thresholds before reaching them, rather than discovering them in front of a client who was expecting a position.
A cell reading « Belgian SME, HR tech, forty staff » identifies the company when it holds two files, however anonymous the number. Five is the usual floor in public statistics, which suppresses cells below it.
Quartiles only, and coarse ones. Thirty is the classical approximation threshold; below it a distribution describes our sample, not a population.
Ten observations per decile. A decile on thirty files is noise presented as a measurement.
Until thirty, one form of publication stays honest: the raw count. « Of twelve surveys, nine had no named authority » is true and checkable. « You are in the third decile » is not.
Get your score
Free instant assessment. Paid reports include public listing.