--- title: "Method — Global AI Monitor" canonical: https://artificial-intelligence.gi/methodology/ source_html: /methodology/index.html description: "How Global AI Monitor jurisdiction records are built: modules, dates, evidence, review labels and limits." generated_at: 2026-11-07T14:18:21Z content_sha256: c8184586b595b90593d4aeb0389def094edaddc576cbbeaf403fbbbfd84d60f2 generator: "fe-fleet/md-twin@1.0" --- # Method — Global AI Monitor Method # How the record is built ## What a jurisdiction page holds Each jurisdiction has one research record. The page for it is generated from that record and shows the same 12 modules in the same order for every jurisdiction: AI Framework Law, Prohibited Practices, High-Risk Systems, General-Purpose AI, Transparency & Synthetic Content, AI & Intellectual Property, Sectoral & Public-Sector AI, Safety & Incident Reporting, Standards & Conformity, Authorities & Enforcement, Sandboxes & Transitional Relief, Commercial Activity. A module with nothing recorded is shown as quiet, with a sentence saying so, rather than hidden. Within a module the page lists the instruments (laws, regulations, guidance and bills), the obligations they create, the milestones they set, and the claims the research makes, each with the evidence behind it. Authorities, enforcement actions, cases and consultations are listed once per jurisdiction. ## Three dates - Research as of: the date of the research run the record comes from. - Record last changed: the last time the record itself was edited. - Published: when the publishing stage has stamped the record. Until then the page says “pending”; in this preview every jurisdiction is pending. Each page also shows the date it was built. A built page is not a newer record. ## Status, evidence and review - Instrument status is stated as at the research date. Superseded and ended instruments are never shown as current. - A milestone whose date has passed is relabelled rather than left as upcoming. - Each claim carries a confidence label and links to its evidence; each piece of evidence names its source, the source tier (T1 is a primary official source; higher numbers are further from the primary text) and whether the source was retrieved. - A claim flagged for human review is shown with the label Under review. It is not hidden, and the label does not mean a person has checked it. - Every count on a page is computed from the record by the page model; none is typed by hand. ## Regulator The regulator line names the body with cross-sector responsibility for AI where one exists. Where none exists the page says so as a declared absence and names who holds the relevant powers. ## AI-content disclosure This service is produced by an AI-assisted research and publishing pipeline from cited primary and authoritative public sources (statutes, regulations, regulator guidance, registers, and official consultations). Every published claim is linked to its underlying source tier and retrieval date; where the pipeline cannot verify a point, the record states that absence explicitly rather than estimating. Published for regulatory intelligence and decision support — not legal, compliance, or investment advice. Editor: Peter Pink-Howitt. Published by Asymmetric Intelligence. Global AI Monitor is assembled by an AI research pipeline from tiered public sources; the editor named below sets its methodology, not the content of each claim. Machine-readable provenance is carried in the page’s ai-generated meta tag ( ), and it is not legal, compliance or investment advice. Facts are machine-assembled from cited public sources. Individual claims have not each been independently verified by a human reviewer. The pipeline roles are Research, Interpreter and Publisher. Where the pipeline has produced nothing, the page says so rather than substituting a generic disclaimer or a plausible estimate. - Silence when nothing material. A module with no findings is shown as an explicit empty state, never filled with invented content. - No fabricated counts. A count or date is shown only where the record backs it. - Cited provenance. The module list, the jurisdiction list and the regulator line come from the site’s configuration and research seed, not invented for the page. ## Limits This is a preview. It covers European Economic Area, United Kingdom, United States – California only. It is decision-support information, not legal, compliance or investment advice, and it is not a substitute for reading the instruments themselves. ## Cost Free to use while coverage is built out. ## Editor and publisher Editor: Peter Pink-Howitt. Published by [Asymmetric Intelligence](https://a-i.gi/), Gibraltar. ## Research benchmark — AI Run window: 16 to 19 October 2026 · Caveats: run by us, AI-judged, not independently validated . Each cell compares stored pipeline research for Global AI Regulatory Monitor (AI) against a single-prompt Google Gemini Deep Research Max report on the same jurisdiction and sector. Figures are ours versus the report. Topic benchmark (AI): Not yet tested in the head-to-head topic benchmark. TRACE dataset audit: Not yet tested in TRACE run-1. Full per-consumer tab and methodology on Asymmetric Intelligence: [https://a-i.gi/benchmark/#tab-ai](https://a-i.gi/benchmark/#tab-ai) · [summary.json](https://a-i.gi/benchmark/summary.json).