v1.3 — The LRRS version

What this is

Version 1.3 changes only the Legal Rail Readiness Score (LRRS), the index's Legal component. It ships seven rulings (#12–#18), adopted on 2026-10-14 after the October 2026 sweep raised questions the methodology had not settled: from which date a legal finding counts, whether a sandbox needs current operating evidence, when a general sandbox counts, whether a law that has not yet commenced counts, and which jurisdictions the survey covers. No other version goes live on the same day, so every change in the score below has one cause.

On the update date (2026-10-20) the live LRRS moves from 13.13 (integer 13) to 10.57 (integer 11). The history from launch is restated on the new rules, with every status on its legal date: at launch, 12.27 as published becomes 10.01. The series as published is kept.

What changed, and why

• Two dates for every finding (Ruling #12). Each status now carries its legal date — the day the fact took effect in the official record — and its update date, the day the verified table containing it went live. The live score counts a finding from its update date; a legal date already in the table (a commencement, an end, a lapse) applies on its own day. At every version that changes an LRRS input, history from launch is recomputed with every status on its legal date. Two cadences, each enforced by a daily check: every stored row is re-verified at least every 35 days, and every cell of the survey grid is searched for new instruments at least every 95 days.
• Operating evidence must be current (Ruling #13). A sandbox scores 1.0 while a participant's test is live, or while its newest operating artifact — an admission, a report on a participant under test, an exit report — is less than 12 months old. When that lapses, the cell falls to 0.5 while the regime stands. France's CNIL sandbox, whose newest artifact is dated 2025-04-18, is 0.5.
• One admitted AI participant qualifies a general sandbox (Ruling #14). A financial or sectoral sandbox counts once the authority's own record names an admitted participant whose system under test uses AI. Australia, South Korea, Saudi Arabia, Japan, India and the financial sandboxes of Spain and Italy enter on that basis, each at the status its current evidence supports.
• A law counts from the day it starts to govern (Ruling #15). An enacted law whose operative provisions have not commenced scores 0, whatever its date. The US GENIUS Act moves from 0.5 to 0 until it takes effect on 2027-01-18; that date is recorded, so it returns to 0.5 on the day. A power to establish a sandbox (“may establish”) is not a regime.
• A stablecoin framework counts on either limb (Ruling #16). Rules for issuers or rules for providers of services in fiat-referenced tokens both qualify, if they address those tokens as such. Brazil's central-bank rules count at 0.5 from 2026-02-02.
• Automated vehicles and drones (Ruling #17). A bounded test of an autonomous machine is judged in the sandbox category by the same four conditions; a standing licence to operate one fits no category. No status changes in this version.
• Every EU member is surveyed (Ruling #18). The survey list becomes the 48 jurisdictions of the GDP table. The LRRS does not change by construction; the survey coverage and its bound do.

The effect on the LRRS

The restated LRRS against the series as published, at representative dates, split in a fixed order: first the data revision (the new table on the old rules, every finding on its legal date), then each rule change.

Ruling #15 is the largest item: under the old rules an enacted law counted from enactment, so the commencement rule removes the GENIUS Act (and the UK's stablecoin regulations, which the data revision adds from enactment) until they commence. Ruling #18 moves nothing by construction.

The effect on the MEI

The LRRS is the MEI's Legal component, one of four equally weighted components. At the same dates, the headline MEI as published, then as each step changes only its Legal component — everything else is the published score's own metric values — and the restated MEI. On the update date it is the first 1.3 score computed from that day's inputs.

The MEI is reported to one decimal and its components as integers, so a step of a tenth or less can round either way. The series of daily scores from launch is restated the same way, each day from its own inputs; the published series is kept.

Every finding

Each status change from the October 2026 table to the 1.3 table, with its legal date, update date, lag and cause. A lag counts from launch for a fact that predates it. The bound is 35 days for an instrument already in the table and 95 days for a new one.

The live change at the update, 13.13 → 10.57, by cause:

Causes are attributed in the order shown, so a finding that leaves its cell unchanged contributes 0: Spain's decree ended, but by the update the cell is held at 1.0 by its financial sandbox. Spain's financial sandbox is dated from 2026-09-29, the earliest official record of its AI participant under test; an earlier official date can move it at the next restatement. The laws Greece, Germany, Malta, Hungary and Poland put in force in July and August entered the live score at the October 2026 update, 63–85 days after they took effect — beyond the 30-day bound published then (cause ii); Slovenia's law, Hong Kong's and Brazil's sandboxes predated launch (cause iii).

Four legal dates are recorded as lower bounds, and all four fall before launch, so none changes a value: Spain's RD 817/2023 from 2025-04-01 (testing began in April 2025; no day was published); the EU (MiCA) and UAE stablecoin rows from the dates their regimes came into force, since the first authorisations were not opened at their source; and South Korea from 2026-05-06, the earliest AI approval opened at its source (an earlier one, of 2024-10-17, was not).

The restated path

The LRRS from launch to the update date on the 1.3 table, each status on its legal date, beside the series as published.

Dates already recorded

These dates are on the table's rows, so each applies on its day with no sweep needed. The monthly re-verification confirms each one in the 35 days before it, and a newer artifact moves a lapse date later.

Survey coverage

The survey list is now 48 jurisdictions — the top 25 economies by nominal GDP, all 27 EU members, and Singapore, the UAE, Hong Kong and the Marshall Islands — covering 89.1% of world GDP on the frozen WEO-2026-04 vintage. If every jurisdiction outside it had full operating coverage in every category, the LRRS would be understated by at most 10.9 points. The realistic understatement is far smaller.

Correction. Since launch the methodology stated a survey coverage of about 87–88% and a bound of 12.5 points. On the frozen vintage the 30 jurisdictions then on the list covered 85.6%, a bound of 14.4 points. The score was not affected: it has always been computed against world GDP.

The coverage log

Every cell of the survey grid — 48 jurisdictions × 5 categories, 240 cells — is searched for new instruments at least every 95 days, and each search is recorded in a published coverage log: the cell, the date, the scope classes covered, the official sources searched, and what was found. A cell's last complete search is its staleness flag, and a daily check fails when one is more than 95 days old. The first complete survey of all 240 cells runs within 95 days of this version; until a cell has one, the log lists it as not yet searched at full scope. The log is at /api/v1/lrrs/coverage-log and on the methodology page.

What did not change, and what is kept

The five categories and their weights, the status ladder's three values, the GDP shares (IMF WEO, April 2026, frozen) and the aggregation are unchanged. Every LRRS table that has been live is kept, and the series as published under version 1.2 is archived (mei_scores_v1_2_archived) and kept. Stored history changes only at a version, in the open.