State of AI Governance 2026

Global coverage and unit of analysis

Coverage is measured against 195 sovereign states: 193 United Nations member states, the Holy See and the State of Palestine. A country result refers to the sovereign-state jurisdiction represented in the country registry. The European Union is tracked as a supranational jurisdiction because its institutions and legislation affect multiple states, but it is not inserted into the country ranking.

Country and government profiles establish the coverage frame. They do not imply that a verified observation exists for every Index indicator. The report therefore presents corpus coverage separately from score publication status.

Institutions and implementation capacity

Institutional evidence is classified by function: government, regulator, standards body, university, research institute and other organization. The presence of an institution is not treated as proof of effective implementation. Records must identify a verifiable mandate, jurisdiction and official source before they can support an Index observation.

Implementation evidence is evaluated independently from policy announcements. Operational responsibilities, oversight, monitoring, enforcement and remedy mechanisms require their own attributable evidence.

Evidence quality and provenance

Every published observation retains its source URL, source date, verification status and review history. Tier indicates source authority, not whether a claim has been independently confirmed. First-party company documentation is therefore attributable Tier 3 evidence and remains distinguishable from regulator findings and independent evaluation.

Archived source versions provide an immutable documentary record for reviewed observations. When a source changes, the replacement is collected as a new version rather than silently overwriting the evidence used in an earlier review.

Rules, strategies and implementation

The knowledge graph separates legislation, regulations, executive instruments, policies, standards and treaty frameworks. Proposed, adopted and in-force instruments are not combined into a single legal-status category. Effective dates and compliance deadlines remain attached to the source that establishes them.

A national AI strategy can evidence policy direction, while binding duties, funded implementation, supervisory activity and accessible remedies are different observations. This separation prevents an announcement from being counted as completed institutional capacity.

How to read Index publication status

A published Index result requires sufficient reviewed evidence across the nine dimensions. Completeness and confidence are disclosed alongside the score so that readers can distinguish a broad evidence base from a narrow one. Equal displayed scores share a rank under the published tie rule.

Until the publication gate is met, the corpus may show verified country facts without displaying a country score or rank. That state is intentional and should not be interpreted as a score of zero or as evidence that governance is absent.

Limitations and update cycle

The source cutoff fixes the evidence window for the 2026 edition, while the live knowledge graph can continue to receive later records. Later legal amendments or newly archived documents do not retroactively change a published edition without a documented revision.

Cross-country comparability is constrained by differences in legal systems, public documentation and implementation transparency. The method addresses these constraints through explicit evidence criteria, completeness reporting and human review rather than by inferring undocumented conditions.

Charts and data

Methodology

The report uses the AI Governance World Index 2026 methodology: nine equally weighted dimensions, forty-five source-traceable indicators and a 31 July 2026 evidence cutoff. Evidence is accepted from official Tier 1 sources, academic Tier 2 sources and attributable industry Tier 3 sources. Media sources are not used as primary evidence.

Each observation retains its source URL, source date, archived source version, verification status and review history. Quantitative observations and documentary rubrics are processed under the published Index rules. Missing observations are excluded from the numerator and reduce completeness; they are not assigned a value of zero. Rankings remain unpublished until the 195-country coverage gate and the minimum completeness requirement are satisfied.

References

1. Regulation (EU) 2024/1689 — Artificial Intelligence ActEuropean Union · Tier 1 · Jul 12, 2024
2. The AI Basic Act Comes into ForceMinistry of Science and ICT, Republic of Korea · Tier 1 · Jan 22, 2026
3. Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of LawCouncil of Europe · Tier 1 · Sep 5, 2024
4. Artificial Intelligence Risk Management Framework (AI RMF 1.0)National Institute of Standards and Technology · Tier 1 · Jan 26, 2023
5. OECD AI PrinciplesOrganisation for Economic Co-operation and Development · Tier 1 · May 3, 2024
6. Recommendation on the Ethics of Artificial IntelligenceUNESCO · Tier 1 · Nov 23, 2021
7. White House Releases New Policies on Federal Agency AI Use and ProcurementThe White House · Tier 1 · Apr 7, 2025