Sources & methodology

This page documents how kya.fi decides what counts as a fact. It exists so that readers, editors and citing authors can judge the reliability of anything we publish — and so that we can be held to a stated method rather than an unstated one.

Our governing rule is simple: sources or silence. If a claim cannot be traced to a real, checkable source, we do not publish it. We never invent facts, quotes, numbers, standards, or vendor claims.

The three-source rule

For any contested, quantitative, or regulatory claim, we require at least three independent primary sources. A parenthetical example: a claim that a standard defines a term a certain way should be supported by the standard itself plus two further independent confirmations (for instance, an implementing guideline and an official explainer from a different authority).

If a claim meets one or two sources but not three, we either label it as reported or uncertain and attribute it to the source we have, or we leave it out.

Descriptive, definitional and clearly labelled analysis statements sit outside the three-source rule; contested claims do not.

Primary sources first

We prefer primary sources over secondary reporting. Where we must use secondary material, we name it as such.

Categories of primary sources we use

  • Standards bodies — ISO/IEC, NIST (NIST SP 800-63-4, NIST AI 100-1), W3C, IETF (RFCs and Internet-Drafts such as `draft-klrc-aiagent-auth` and `draft-oauth-ai-agents-on-behalf-of-user`).
  • Regulators and official guidance — the European Commission’s AI Act service desk, the European Data Protection Board, national data-protection authorities, and comparable bodies.
  • Vendor documentation — official, first-party docs from the organisation that owns the system, cited as that organisation’s own claim (never as neutral fact).
  • Security research and vulnerability data — maintainer advisories, CVE records, and reproducible or peer-reviewed research; we say plainly when a finding is disputed. The gap these sources do not by themselves cover — why an agent is not just another workload — is set out in Why agent identity is a new problem.

Source ledger format

Each non-obvious claim is backed by a ledger entry so it can be audited. The format is:

FieldMeaning
ClaimThe exact statement being supported.
Status`verified` (three-source), `reported` (fewer sources), `opinion`, or `uncertain`.
Source titleThe title of the work.
PublisherThe organisation that issued it.
URLA direct, stable link.
AccessedThe date we last confirmed the link and content (e.g. `2026-10-08`).
Supporting quoteThe shortest passage from the source that supports the claim.

A worked example:

ClaimStatusSource titlePublisherURLAccessedSupporting quote
GDPR gives data subjects the right of accessverifiedRegulation (EU) 2016/679, Article 15European Union (EUR-Lex)https://eur-lex.europa.eu/eli/reg/2016/679/oj2026-10-08“Right of access by the data subject”

Review cadence

  • Every publish — the source ledger is completed and every URL is checked before a page goes live.
  • Standards and regulation pages — revisited at least quarterly, because drafts, timelines and regulator guidance change.
  • All content — reviewed at least annually for accuracy and dead links; time-sensitive statements carry an “as of” date.

Suggesting a source or flagging a claim

We welcome correction. Email editor@kya.fi or use Contact with:

  1. the page URL or title,
  2. the specific claim you are querying, and
  3. a primary source, if you have one.

We aim to acknowledge substantive reports and will update the page — and, where relevant, note the change — when a correction is warranted.

Primary sources


Know Your Agent (KYA) explains agent identity, verification and accountability. This is an explainer, not legal or compliance advice — see our Sources & methodology. New to KYA? Start with What is Know Your Agent? and the glossary.