規制データ標準と開示されたAI志向：米国Inline XBRL段階導入からの証拠

Regulated Data Standards and Disclosed AI Orientation: Evidence from the US Inline XBRL Phase-In (原題)

Alessio Faccia

Journal of Risk and Financial Management📚 査読済 / ジャーナル2026-10-03#AI×ESGOrigin: US対象セクター: finance

DOI: 10.3390/jrfm19100766

原典: https://doi.org/10.3390/jrfm19100766

🤖 gxceed AI 要約

日本語

米国Inline XBRLの段階導入を250社・1450件の10-Kで検証。規制データ標準の導入がAI関連開示の広がりと関連するかを分析した。Holm補正後、4つの開示指標はいずれも仮説を支持せず、AI志向の調整効果も有意でなかった。開示の広さと実証された運用能力は区別すべきと結論づける。

English

Using 1,450 Form 10-K firm-years from 250 CIKs (FY2014–2024), this study examines whether the US Inline XBRL phase-in is associated with disclosed AI orientation. After Holm adjustment, none of four reporting indicators support the hypothesis, and AI-related disclosure breadth shows no moderation. It distinguishes disclosure breadth from verified operational capability.

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

📝 gxceed 編集解説 — Why this matters

日本のGX文脈において

SSBJ基準・有報のXBRL化・統合報告のデジタル開示インフラ整備を進める日本にとって、規制データ標準が企業のAI・無形資産開示に与える影響を実証的に示す貴重な先行例。開示の広さと実態能力の乖離は日本企業の情報開示設計にも示唆を与える。

In the global GX context

As ISSB/CSRD push machine-readable disclosure (XBRL/ESEF), this paper offers rare empirical evidence on whether regulated data standards shift firm-side AI and intangible disclosure. Its null results caution against equating disclosure breadth with verified capability—relevant to global disclosure-infrastructure debates.

👥 読者別の含意

🔬研究者:規制データ標準と企業開示行動の因果推論における非ランダム導入・左打ち切り問題の扱い方を学べる。

🏢実務担当者:XBRL対応がAI・無形資産開示の広さを必ずしも増やさない点を踏まえ、開示の質と実態の整合を意識すべき。

🏛政策担当者:機械可読開示の義務化が企業のAI関連開示を促すという想定には慎重な検証が必要だと示唆する。

📄 Abstract（原文）

Machine-readable reporting standards provide shared data infrastructure, while firm-side reporting associations may vary with technological disclosure. The study examines the US Inline XBRL phase-in using 1450 Form 10-K firm years from 250 CIKs over fiscal years 2014–2024. SEC filing metadata, XBRL facts, amendments, accounting controls and full-primary-document text form the analytical panel. Preferred models date exposure from the first periodic report whose SEC metadata identify Inline XBRL use and include firm and year fixed effects with firm-clustered inference. Observed entry is non-random, 41 firms are left-censored for event-time analysis, 82 firms lack observed entry, and later periods offer limited untreated support. The design therefore estimates conditional associations, not a mandate effect. None of the four reporting indicators support H1 after Holm adjustment at the 5 per cent level. Dictionary-defined AI-related disclosure breadth, described as disclosed AI orientation, yields no preferred moderation evidence, and the intangible-intensity and innovation-disclosure tests also remain insignificant after adjustment. A fully specified fixed pre-entry moderator produces two suggestive interactions with Holm-adjusted p = 0.080, without rejection at 5 per cent. Extending asset-threshold and control-omission tests across the main outcomes does not overturn that threshold-based inference. Custom-tag leads reject a flat pre-entry path. The findings distinguish disclosure breadth from verified operational capability and quantify remaining uncertainties rather than establish exact zero.

🔗 Provenance — このレコードを発見したソース

semanticscholar https://doi.org/10.3390/jrfm19100766first seen 2026-10-08 05:47:24 · last seen 2026-10-10 05:57:52

🔔 こうした論文の新着を逃したくない方はキーワードアラートに登録（無料・3キーワードまで）。

gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。