The plumbing is there; the public record is not
A visibility paradox
The EU-27 score highest on technical trust (92/100 — identity, PKI, cybersecurity baselines) but lowest on public inventory (42/100). States have the trust machinery; few publish the registers that would let anyone see the algorithms in use.
Visibility is uneven
Of 27 member states, only 3 reach class A (visible governability); 13 are structured-but-incomplete (B), 8 fragmented-but-observable (C), and 3 show only thin visibility (D). The map is lumpy, not uniform.
The agentic blind spot
The index measures whether the state itself could trace AI-agent action from legal basis to lifecycle evidence. Even the strongest member states leave the agent layer — who acted, under whose authority, on what evidence — largely outside the public record.
Where observability concentrates
Observability across the EU-27
| # | Member state | Score | Class | Law/auth | Pub. inv. | Tech. trust |
|---|---|---|---|---|---|---|
| 1 | France | 76.7 | A | 67 | 67 | 100 |
| 2 | Netherlands | 76.7 | A | 67 | 87 | 83 |
| 3 | Slovenia | 76.7 | A | 80 | 67 | 100 |
| 4 | Germany | 71.7 | B | 67 | 47 | 100 |
| 5 | Spain | 71.7 | B | 67 | 67 | 83 |
| 6 | Greece | 68.3 | B | 67 | 67 | 100 |
| 7 | Denmark | 66.7 | B | 67 | 47 | 100 |
| 8 | Portugal | 66.7 | B | 47 | 47 | 100 |
| 9 | Sweden | 66.7 | B | 67 | 47 | 100 |
| 10 | Estonia | 65.0 | B | 53 | 47 | 100 |
| 11 | Finland | 65.0 | B | 53 | 67 | 100 |
| 12 | Czechia | 63.3 | B | 67 | 47 | 100 |
| 13 | Italy | 63.3 | B | 67 | 47 | 100 |
| 14 | Poland | 61.7 | B | 67 | 33 | 100 |
| 15 | Austria | 60.0 | B | 67 | 47 | 83 |
| 16 | Luxembourg | 60.0 | B | 67 | 27 | 100 |
| 17 | Lithuania | 58.3 | C | 47 | 47 | 100 |
| 18 | Slovakia | 55.0 | C | 67 | 13 | 100 |
| 19 | Belgium | 53.3 | C | 33 | 47 | 100 |
| 20 | Malta | 53.3 | C | 67 | 13 | 67 |
| 21 | Latvia | 50.0 | C | 33 | 27 | 100 |
| 22 | Ireland | 48.3 | C | 33 | 47 | 67 |
| 23 | Croatia | 45.0 | C | 33 | 33 | 83 |
| 24 | Hungary | 45.0 | C | 33 | 13 | 100 |
| 25 | Romania | 41.7 | D | 33 | 13 | 83 |
| 26 | Cyprus | 38.3 | D | 33 | 13 | 67 |
| 27 | Bulgaria | 36.7 | D | 33 | 13 | 67 |
| # | Top 10 worldwide | Score | Cl. |
|---|---|---|---|
| 1 | Singapore | 81.7 | A |
| 2 | Japan | 80.0 | A |
| 3 | Norway | 78.3 | A |
| 4 | United Kingdom | 78.3 | A |
| 5 | China | 76.7 | A |
| 6 | France | 76.7 | A |
| 7 | Netherlands | 76.7 | A |
| 8 | Slovenia | 76.7 | A |
| 9 | Germany | 71.7 | B |
| 10 | Spain | 71.7 | B |
This observatory reports the visibility of AI-governance machinery in public records, scored on the frozen 10-indicator indicator matrix (each indicator 0–3, weighted to 100; a mild H/M/L confidence adjustment). It does not rate compliance, make a supervisory finding, determine legal status, or audit institutions — weak public evidence is a visibility signal, not proof that machinery is absent. Scores follow each jurisdiction's public record at a snapshot date and are never fused across surfaces. Codebook and dataset: Zenodo, CC-BY.