How this register is built

The register lists in-force sustainability, environmental and energy regulation across 119 jurisdictions, synced weekly from the XG corpus, each row with a link to its official source. It excludes repealed or superseded instruments and proposals that have no adopted official text. A row without a traceable primary source does not enter the corpus at all.

Those links are not all equally deep. 679 of 850 rows — 80% — link to the instrument or the gazette notice itself. The remaining 171 link to the ministry, regulator or standard-setter that publishes it, which means finding the text takes one more step. Improving those is corpus work, and until it is done this is the honest description of what the column gives you.

It is a working corpus, not an exhaustive statute book: 850 rows is what XG tracks and can source, not every environmental law in every country. The full index carries the same scope statement above its table.

Row-level facts — names, jurisdictions, statuses, effective dates, official source links — come from the XG corpus, a database XG maintains for its consulting work. A scheduled job runs every Monday at 06:00 UTC and syncs the corpus into this site's data file. When the data has actually changed, the job opens a pull request; a person reads the diff and merges it. Merging is the review gate, deliberately — nothing lands on public pages about legal obligations without a human looking at what moved. An unchanged corpus produces no pull request at all.

What that job does not do is check the law. It reads the corpus and nothing else — it does not query gazettes, regulators or legal databases, and it cannot notice a change the corpus has not already recorded. Keeping the corpus current is manual work done by people, prompted by a regulator's announcement or a reader's correction rather than by a schedule. So a week in which the register does not move means the corpus did not move; it is not evidence that nothing changed in law.

The sync moves facts only, never text. Corpus prose is deliberately not imported — the prose on written-up pages is written and verified by hand, and the sync cannot overwrite it. The sync output is deterministic, so any diff in the weekly pull request means the corpus itself changed.

Most of the 850 rows are exactly that: a row, with a link to the official source. A regulation earns a written-up page only when two things hold — the register has a penalty or threshold for it verified against a primary source, and there is a demand signal on record, such as search impressions or a client question. Never before: the register publishes no stub profiles, and a row without a page links to the instrument itself rather than to a placeholder.

Written-up pages are additionally released in reviewed waves: a wave of pages is drafted, fact-checked against live sources, corrected, and only then published. 21 pages are currently live under this rule.

Every penalty figure in the register carries one of three confidence grades, and the grade is printed next to the figure — never colour-coded away.

• High

  The figure is traced to the primary source itself — the gazette or official text. A high-confidence figure renders with no mark at all: silence means verified.

• Medium

  The figure is marked "reported, not gazette-verified": it is confirmed across consistent independent analyses of the enacted text, but has not been line-read in the gazette itself — usually because the gazette could not be machine-read.

• Unverified

  Marked "no figure traced to a primary source". The register holds no amount it can stand behind, and says so in words rather than omitting the row.

"No figure traced to a primary source" means exactly that: the member state may well enforce, but no maximum could be established from a gazette or competent authority. The register publishes that absence as a row rather than filling it with a number from secondary commentary. Figures circulate widely that do not survive verification — five amounts previously held in the corpus were rejected on checking against primary sources and are deliberately absent from the CSRD dataset. Currently 17 of the 27 member states in that dataset carry an honest blank.

Every written-up page carries a corrections form — readers of regulatory pages are often the first to know a law moved. A report names the page, the field it concerns, what is wrong, and optionally a source. Each report becomes a review ticket and is reviewed alongside the corpus sync; reports are read and checked by a person, never applied automatically. When the reporting endpoint is unavailable, the form falls back to an email address rather than pretending the report was filed.

Written-up pages are reviewed by XG Energy's regulatory research team — an organisation byline, stated as such. "Last verified" on a page is the date its stated positions were checked against the primary sources listed on that page; it is not the build date, and it is distinct from "data last changed", which records when the underlying data last moved — the most recent changelog entry or the corpus row's own update date, whichever is fresher.

Coverage is not exhaustive, and depth is uneven: some jurisdictions hold a single row. The corpus can lag a legal change indefinitely — the weekly sync mirrors the corpus rather than monitoring the law, so a change nobody has recorded stays unrecorded — and a fact change on a written-up page can briefly outrun its prose, which is corrected at review.

Two CSRD rows still carry an explicit human-check caveat, printed on the pages themselves. Austria: the gazette for the enacted NaBeG returned a server error on every route on 9 August 2026 and again on 10 August 2026, so the penalty tiers rest on four independent Austrian analyses of the enacted text, which agree — the row stays at medium confidence because nobody here has read the gazette itself. Ireland: the Irish Statute Book returned an access error, so the offence-category maxima follow the companies registry’s published leaflet.

A third caveat was closed on 10 August 2026, and closing it changed a published number. Slovakia’s sustainability-reporting fine had been shown as €3,000,000, then corrected to €1,000,000 on professional analyses. Reading the consolidated Accounting Act itself showed the tier that penalises the sustainability offence is €100 to €100,000 — an order of magnitude lower. The larger figure belongs to a neighbouring provision covering assurance and filing failures, and the secondary sources had attached it to the wrong offence. The row now cites the statute and is at high confidence. It is the clearest evidence this register can offer for its own rule: sources that agree with each other can still be wrong together, and only the primary text settles it.

The CSRD penalty dataset is a verified snapshot regenerated when the underlying data changes, not a live query — its verification date is printed on every page that uses it. The register's own statistics are computed from these same files at every build; they are published at ESG regulation statistics.