Every AI Policy Ever · open in the interactive register

Published (non-binding, current)Non-binding · Guidelines / framework

Guidelines on Assessment of AI Reliability in the Field of Plant Safety

プラント保安分野AI信頼性評価ガイドライン

Record JP-15

Issued by
Liaison Council of Three Ministries on Disaster Prevention of Petroleum Complexes (Ministry of Economy, Trade and Industry; Fire and Disaster Management Agency; Ministry of Health, Labour and Welfare)
Jurisdiction
Japan
Level · region
National · East Asia
Published / adopted
2020-11
Last updated
2021-03
Language
Japanese, English

Summary

Guidelines of the liaison council of METI, the Fire and Disaster Management Agency and the Ministry of Health, Labour and Welfare on assessing the reliability of machine learning used in petroleum and chemical plants, where safety is crucial. Following AIST's Machine Learning Quality Management Guideline, they manage reliability through quality in use, external quality and internal quality, with required levels for risk avoidance and performance, and take account of external safety mechanisms, safety systems independent of the AI and human involvement in its decisions. They apply this to five plant use cases and set out tasks in development and operation.

Safety topics

critical infrastructureevaluation & testinghuman oversight

Status notes

The first edition was published in November 2020. The second edition of March 2021 clarified how AI safety assessment relates to existing plant risk assessment methods such as HAZOP and FMEA, added explanations to the template for recording reliability assessments, a Q&A and seven practical examples covering the five use cases (pipe wall thickness prediction, pipe image diagnosis, equipment deterioration diagnosis, detection of early signs of abnormality, operation optimisation). The guidelines were drawn up from a working group of plant operators and safety and AI experts that met from April 2020 to March 2021, and are to be reviewed as they are used. The council publishes an English edition of the second edition.

Sources

Open the official source https://www.fdma.go.jp/relocation/neuter/topics/fieldList4_16/jisyuhoan_shiryo.html

Local copy

Archived. Last attempt 2026-10-18 00:46:11 UTC

Official source (third-party copy, not published) archive/_runs/2026-10-14-discover-leads-2026-10-14/C020/2026-10-14/official.html
Retrieved 2026-10-14 13:11:38 UTC
SHA-256 c60df78e0244445c0e44253681cf40c489feec561e2f450c4d68b718a315f8c0
Linked document (PDF) (third-party copy, not published) archive/_runs/2026-10-14-discover-leads-2026-10-14/C020/2026-10-14/document-1.pdf
Retrieved 2026-10-14 13:11:42 UTC
SHA-256 d5450d1d743400dde20ac66a489a52955a221863324d23730d1f7f812f5552cc
Linked document (PDF) (third-party copy, not published) archive/JP-15/2026-10-18/document-2.pdf
Retrieved 2026-10-18 00:14:22 UTC
SHA-256 1c27c2146dcc51aa32809558b00b79796002fc380b5de0b8c29f1ba6ed194364

Verification

A
The instrument's own official text, or an official register, was stored and confirms the title, date and status. checked 2026-10-18 · official-url-fetched
Recheck of 18 Oct 2026. Stored and read: the three-ministry council's page on the FDMA site (fdma.go.jp; copy of 14 Oct 2026, unchanged on 18 Oct 2026); the Japanese second edition (151 pages, 2021 年 3 月, 石油コンビナート等災害防止３省連絡会議; copy of 18 Oct 2026); the council's English edition of the second edition (158 pages, March 2021, with the working group's members in the annex); and the Japanese summary of the revision (copy of 18 Oct 2026, identical to that of 14 Oct 2026). They confirm the title, the Japanese title, the issuers, the first publication in November 2020 (2020年11月; 令和2年11月の公表以来), the second edition of March 2021, the study group of April 2020 to March 2021, the intention to keep reviewing the guidelines and the content; the summary is written from the English edition, which the council publishes as its English edition without a translation notice, and checked against the Japanese text. The archive's flag that a source changed on 18 Oct 2026 does not concern this instrument: that day the archiver stored, in the record's document-2 slot, a different PDF linked from the page (r05/jisyuhoan_shiryo_01.pdf, a summary of a survey on non-explosion-proof equipment in plants), not a new version of the revision summary. Nothing stored shows an edition after March 2021 or a withdrawal.
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 7 field(s) supported.

History

  1. 2026-10-18Re-verified against its sources, no change — pipeline run 2026-10-18-recheck-east-asia-1 (recheck), candidate JP-152026-10-18-recheck-east-asia-1 · Claude Code (claude-opus-5-5)
  2. 2026-10-14Added to the register — found in pipeline run 2026-10-14-discover-leads-2026-10-14, candidate C020 (JP/ministries)2026-10-14-discover-leads-2026-10-14 · Claude Code (claude-opus-5-5)

Citation

Liaison Council of Three Ministries on Disaster Prevention of Petroleum Complexes (Ministry of Economy, Trade and Industry; Fire and Disaster Management Agency; Ministry of Health, Labour and Welfare). Guidelines on Assessment of AI Reliability in the Field of Plant Safety. 2020-11. https://www.fdma.go.jp/relocation/neuter/topics/fieldList4_16/jisyuhoan_shiryo.html. Status: Published (non-binding, current); verified 2026-10-18 (level A). Local copy archive/_runs/2026-10-14-discover-leads-2026-10-14/C020/2026-10-14/official.html retrieved 2026-10-14 13:11:38 UTC, SHA-256 c60df78e0244445c0e44253681cf40c489feec561e2f450c4d68b718a315f8c0. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record JP-15.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.