Editorial principles

A document is eligible when it is issued or formally endorsed by a public authority, standards body, or international institution, and when it addresses artificial intelligence directly rather than incidentally.

Drafts, consultations and proposals are included when they are published officially and have a durable public source. Commentary, opinion pieces and vendor material are out of scope.

Every record must link to a primary source: an official gazette, ministry page, parliamentary record, standards catalogue or institutional publication. Secondary reporting is never used as the sole basis for a record.

Metadata — jurisdiction, issuing body, publication date, document type and status — is transcribed from the source document, not inferred, and is left blank rather than guessed when a source is silent.

Records are revisited when a source document changes status, is amended, superseded or withdrawn. The archive keeps the original record and notes the change rather than deleting history.

Each entry carries a last-updated date so researchers can judge how current it is.

Factual errors are corrected promptly and without argument. If you find an inaccurate date, mis-attributed issuing body, broken source link or mistaken summary, report it and it will be reviewed against the primary source.

Substantive corrections to a summary are reflected in the record's last-updated date.

Summaries are descriptive. They state what a document says and does, not whether it is good, sufficient or advisable.

The archive publishes no scores, rankings, league tables or maturity grades. Where comparative views exist, they show presence and absence of recorded material only.

Derived groupings — topics, themes and relationship links — are produced from recorded metadata and document text using published rules, not editorial judgement about importance.

Where the archive infers a relationship between records, it is labelled as derived so researchers can verify it themselves against the sources.

In scope: national AI strategies, binding legislation and regulation, regulatory guidance, ethics frameworks, technical and management standards, public-sector procurement and use policies, and international agreements and declarations.

Out of scope: private company policies, unpublished internal guidance, academic literature, and general digital or data policy that does not address AI specifically.