EasyxLabOpen studies and verification tools for the rules software has to follow.https://easybyte.es/lab/2026-10-03T00:00:00ZEasyxLab · EasyByte Hub S. Coop. Mad.S17 — High-value datasets on data.europa.eu: what a machine sees, and what Member States sayhttps://easybyte.es/lab/studies/s17/2026-10-03T00:00:00Z2026-10-03T00:00:00ZOn 2026-10-03, 28 months after the rules became applicable, the European data portal identified no high-value dataset for 7 of the 27 Member States, 3 of which had told the 2025 Open Data Maturity survey that their portal uses the DCAT-AP HVD tag; another 13,948 records (12,062 distinct datasets) carried an HVD category without the legal reference the portal needs to count them.S11 — Do the registration numbers on Airbnb listings exist in the official registries? Spain and New York around 20 May 2026https://easybyte.es/lab/studies/s11/2026-10-03T00:00:00Z2026-10-03T00:00:00ZIn June 2026, in five Spanish areas with an open registry, 2,066 of 31,222 listings showing a well-formed tourist-dwelling number (6.6%) showed one we could not trace to any registered dwelling; a further 557 (1.8%) showed a registered dwelling written in a recognisable wrong form.S10 — Past the static deadlines: NeTEx on five European access points, checked against open schemashttps://easybyte.es/lab/studies/s10/2026-10-03T00:00:00Z2026-10-03T00:00:00Z8 of 34 timetable datasets from five national access points are fully valid against the NeTEx XSD, all of them French; the static MMTIS deadlines passed in 2019–2025, and the 1 December 2026 date covers parking and vehicle sharing, not timetables.S8 — What can a machine verify on Spain's public-sector websites?https://easybyte.es/lab/studies/s8/2026-10-03T00:00:00Z2026-10-03T00:00:00ZOf 5,130 public-sector home pages served over HTTPS, 1,688 (32.9%) send HSTS; of 5,570 entities, 4 serve a security.txt that is strictly valid under RFC 9116.S7 — Do Google's AI Overviews and AI Mode keep up with Spanish rule changes? About one answer in twenty did nothttps://easybyte.es/lab/studies/s7/2026-10-03T00:00:00Z2026-10-03T00:00:00ZJudged against the consolidated law in force on 2 October 2026, 12 of 240 AI Overview answers (5.0%) and 13 of 291 AI Mode answers (4.5%) on 27 recently changed rules were outdated or wrong.S6 — Do AI-generated images on Wikimedia Commons carry provenance marks?https://easybyte.es/lab/studies/s6/2026-10-03T00:00:00Z2026-10-03T00:00:00ZFiles with a machine-readable AI mark rose from 25.8% of January–May 2026 uploads to 50.2% of June–July uploads; our own validator, ai-mark-lint 0.1.0, wrongly rejected 187 valid C2PA manifests (fixed in 0.1.1).S5 — Who publishes attestations on PyPI and npm — and who stopped?https://easybyte.es/lab/studies/s5/2026-10-03T00:00:00Z2026-10-03T00:00:00ZThe latest version is attested for 3,479 of 14,995 top PyPI projects (23.2%); 138 projects that once attested no longer do in the line pip installs, 96 of them after a change of publishing tool or workflow.S4 — Multilingual quality of drug names in Wikidatahttps://easybyte.es/lab/studies/s4/2026-10-02T00:00:00Z2026-10-03T00:00:00ZAbout 0.9% of 34,207 labels are wrong (a lower bound), concentrated in Urdu, Persian, Russian and Hindi; 135 proposed corrections, not applied.S3 — Do AI provenance marks survive common image, video and audio pipelines?https://easybyte.es/lab/studies/s3/2026-10-02T00:00:00Z2026-10-03T00:00:00ZEvery pixel or container rewrite removed the embedded C2PA manifest (124/124); editing one metadata field after signing left it present but invalid (6/6).S2 — Spoofed AI crawlershttps://easybyte.es/lab/studies/s2/2026-10-02T00:00:00Z2026-10-03T00:00:00Z39.3% of 16,548 claims on three small sites were spoofed; for user-initiated fetchers (ChatGPT-User, Claude-User, Perplexity-User…) 67.8%.S1 — Conformance of open-source Verifactu implementationshttps://easybyte.es/lab/studies/s1/2026-10-02T00:00:00Z2026-10-03T00:00:00ZOf 35 files presented as valid records, 10 (28.6%) contain at least one error; the most frequent defect is a declared hash that does not match the record.