Trust & verification
Claim substantiation
The coverage matrix is 100% primary-source-cited (/wiki/verification). This measures the harder case: the developed editorial prose. Each article’s prose is split into sentences; every sentence making a checkable factual claim should carry an inline citation. We report the rate and list the uncited claims, so the gap is actionable.
Heuristic, deterministic claim-detection — a measured signal, not a proof; it will miss some claims and over-flag some framing. Machine-readable: /wiki/claim-substantiation.json.
Independently cited
65%
Self-pinpoint only
193
Uncited claims
315
Any citation
78%
The headline is independentcitations only. A bare provision pinpoint (Art./§) in an article about its own statute is a self-reference to the cited primary source — legitimate, but not an independent high-quality reference — so it counts toward “any citation” (78%) but not the 65% headline.
Substantiation closure
Measuring the gap is not closing it. An offline adversarial panel (refute-by-default, independently over-crediting-audited) classifies genuine uncited claims by their substantiation route. Of the 150 classified claims, 86% are already substantiated — 62by the article’s own cited primary source (a legitimate route the inline-citation heuristic cannot see), 6 by an in-corpus reference, and 61 by a research + fetch-verified external reference (a real source found and independently confirmed to support the claim). 21 still need a new external reference (the actionable sourcing queue); 15 were framing the detector over-flagged (this over-flag route is not yet independently re-audited). A further 126 claims are not yet classifiedand are excluded from the closure rate — recently-added claims from ongoing citation and prose work, plus entries returned to the queue after an independent re-audit found the “not a claim” route had been drawn too broadly. Surfaced here rather than hidden; closure above is the share of the classified set only. Machine-readable: /wiki/substantiation-ledger.json.
Closure (of classified)
86%
Primary-source-backed
62
Needs a reference
21
Not yet classified
126
Over-flagged (not a claim)
15
Lowest-substantiated articles
929 of 1437 prose factual claims carry an independent citation; 193 rest only on a self-referential provision pinpoint. The articles below have the lowest independent-citation rate — the editorial review queue.
- AI in Criminal Justice5/20 indep. (25%)
e.g. “The EU treats one application — purely profiling-based individual crime prediction — as an unacceptable risk warranting an outright ban (Reg.”
- OMB Memorandum M-24-10 (Advancing Governance, Innovation, and Risk Management for Agency Use of AI)2/6 indep. (33%)
e.g. “Issued March 28, 2024 under the Director of the Office of Management and Budget's authority to bind covered executive agencies, M-24-10 operationalizes Executiv…”
- Designated Systemic-Risk Model3/9 indep. (33%)
e.g. “Once a model crosses 10²⁵ FLOPs, the provider must notify and is presumed systemic; the AI Office may also reach below the threshold using Annex XIII indicators…”
- Environmental Impact of AI Training6/17 indep. (35%)
e.g. “Outside AI-specific law, France regulates the same footprint through general digital-environment rules: the REEN Act (Loi n° 2021-1485) underpins the ARCEP–ADEM…”
- Capability Elicitation3/8 indep. (38%)
e.g. “Elicitation thus reframes evaluation from observed conduct to demonstrated potential.”
- Executive Order 14110 on Safe, Secure, Trustworthy AI10/26 indep. (39%)
e.g. “Reg. 75191 (Nov. 1, 2023)) operated chiefly as a tasking instrument that directed federal agencies to produce binding sub-regulation, rather than imposing dutie…”
- Synthetic Content Provenance5/13 indep. (39%)
e.g. “The binding instruments converge on the same goal — make AI-origin detectable — but split on WHO must act and by what modality.”
- GSA Generative AI and Specialized Computing Infrastructure Acquisition Resource Guide7/18 indep. (39%)
e.g. “Issued April 29, 2024 (GSA, Generative AI and Specialized Computing Infrastructure Acquisition Resource Guide), the Guide operates through procurement plumbing …”
- Council of Europe Framework Convention on AI9/23 indep. (39%)
e.g. “The EU's own ratification (reported 15 May 2026, coe.int) positions the AI Act as the EU's chief instrument of compliance, making the Convention a normative flo…”
- FedRAMP AI Cloud Procurement Guidance6/15 indep. (40%)
e.g. “The guidance also cross-walks to OMB M-24-10 minimum practices for safety- and rights-impacting AI, positioning FedRAMP as the security gate beneath a separate …”
- EU AI Act11/26 indep. (42%)
e.g. “China's regime is binding but vertical and rolled out piecemeal — the Interim Measures for the Management of Generative AI Services (effective 15 August 2023) t…”
- Training-Data Rights11/25 indep. (44%)
e.g. “Copyright Office, May 2025).”