Have Your Say: stakeholder responses on EU digital-identity, trust-service and AI files (2026 snapshot)
收藏资源简介:
A corpus of 2,244 published stakeholder responses across 16 European Commission "Have Your Say" initiatives on EU digital identity, trust services and AI, harvested in August 2026 from the Commission's public feedback interface, with an index of 521 attachments carried by 507 responses. Why. The EU is writing the implementing detail of how automated systems act on behalf of people — the eIDAS 2.0 wallet, the attestation-of-attributes rulebooks, the AI Act's logging and oversight duties, and the Cloud and AI Development Act. This dataset makes the consultation record analysable: who argues about accountability for delegated action, and which arguments are missing. Privacy. The Commission publishes respondents' personal names for "publish with information" responses. This dataset removes them — no name or login field is retained. Organisation names are kept, since the unit of analysis is the respondent class and the named organisation. Free-text fields were additionally scanned and 11 email addresses and 1 telephone number redacted in place. No other alteration was made to response text. Two facts to know before analysing. (1) More than half of the raw corpus is a single national citizen mobilisation: 1,193 responses are from Slovakia, of which 1,190 are citizens opposing digital identity. Exact-text clustering flags only 17 responses as identical templates, because participants vary their own wording — template detection alone will suggest the corpus is clean when 53% of it is one country's campaign. The 786 non-citizen responses are the professional debate. (2) Only published (WITHINFO) responses are visible; anonymous and unpublished responses are absent, so this is not the population of opinion and supports no inference about influence. A worked finding. Among the 786 organisational responses, counted with word-bounded phrase lists (see the analysis script in this version): portability, lock-in and interoperability are engaged by 124; delegation or acting on behalf by 77; attestation of the acting runtime by 4 on the strict list and 15 when secure-hardware vocabulary (HSM, WSCD, remote QSCD) is admitted; audit or certification in the assurance sense by 70; and evidence that survives the action for later verification by 7. Academic or research institutions account for 32 responses in the whole corpus. Concept counts are a keyword floor over response text, so arguments made only inside uploaded attachments are undercounted and these figures are lower bounds. Version 2 (18 August 2026). The response data (JSONL, CSV) is byte-identical to version 1. This version corrects the README and this description — the version-1 attestation count of 62 was an artefact of the bare token "attestation" matching the regulation's own attestation-of-attributes vocabulary rather than runtime attestation (61 organisational responses contain the token; 42 of them use it in the attribute-attestation sense); portability 123→124 and delegation 74→77 reflect extended, word-bounded lists; the initiative composition is 1 eIDAS application report + 13 implementing acts + business wallet + Cloud and AI Development Act, previously miscounted as "eleven" implementing acts. It also adds three files so that every published figure is reproducible from the deposit alone: hys_data_paper_stats.py (standard-library analysis script; regenerates every figure of the accompanying data article from the JSONL), pubs.json (the 43 publication records with consultation window end dates, required for the completeness and deadline-geometry analyses), and hys_corpus.py (the harvester). Interactive observatory over this corpus: https://says.eatf.eu Source data is public content published by the European Commission and its respondents; this is a derived, privacy-reduced compilation. A visibility observation of public data at a snapshot — not a representativeness or influence claim.



