Every AI Policy Ever · open in the interactive register

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

Guidelines on Technical Measures to Ensure the Security of AI

AIのセキュリティ確保のための技術的対策に係るガイドライン

Record JP-12

Issued by
Ministry of Internal Affairs and Communications
Jurisdiction
Japan
Level · region
National · East Asia
Published / adopted
2026-03
Language
Japanese

Summary

Guidelines of the Ministry of Internal Affairs and Communications on technical measures that AI developers and AI providers can take to secure AI, mainly large language models and systems that contain them. Building on the AI Guidelines for Business and J-AISI's evaluation guide, they focus on direct and indirect prompt injection and denial-of-service attacks, and also cover data poisoning, attacks through tampered models and model extraction. Measures include training on safety standards, system prompts, guardrails that check inputs, outputs and external data, and permission management for orchestrators and RAG, with red teaming to confirm that they work. AI agents are excluded for now.

Language

No English version of this instrument was found. The English title and the summary are the register's own, written from the sources below; for the exact wording use the Japanese text.

Safety topics

evaluation & testinggeneral-purpose / frontier AI

Status notes

MIC announced the final guidelines in a press release of 27 March 2026; the guidelines themselves are dated March 2026. A draft was open for public comment from 26 Dec 2025 to 29 Jan 2026 and drew 52 comments; MIC published them with its views alongside the final guidelines and a separate annex of supporting material. The guidelines follow the conclusions, reached in December 2025, of the AI Security Subcommittee that MIC convened from September 2025 under its Cybersecurity Task Force. They state that they do not decide who is responsible for a threat, and MIC intends to supplement them as new threats emerge, for example from vision language models, AI agents and MCP.

Sources

Open the official source https://www.soumu.go.jp/main_content/001064122.pdf

Secondary source: https://www.soumu.go.jp/menu_news/s-news/01cyber01_02000001_00282.html

Local copy

Archived. Last attempt 2026-10-18 07:41:58 UTC

Official source (third-party copy, not published) archive/JP-12/2026-10-18/official.pdf
Retrieved 2026-10-18 07:42:04 UTC
SHA-256 083731c5a8e96b9a64a08a89f03c435ce620ade2260ca8b8447a59a5bd106d8c
Secondary source (third-party copy, not published) archive/JP-12/2026-10-18/secondary.html
Retrieved 2026-10-18 07:42:08 UTC
SHA-256 9b6a3fadf2c3b11629839804844b8dc0516f7ef41d0b9fb42fc93b0444a84eb6

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
Refresh of 18 Oct 2026 under the rule that level A needs the instrument's own text or an official register. Stored and read (soumu.go.jp, stored 14 Oct 2026 and unchanged when the archiver checked them on 18 Oct 2026): the guidelines' main text (25 pages, cover 令和８年３月, 総務省), now the official_url; MIC's Japanese press release of 27 March 2026 (令和8年3月27日), the previous official_url, now the secondary source; and the table of the 52 public comments with MIC's views (18 pages, comment period 令和７年12月26日 to 令和８年１月29日). The guidelines confirm the Japanese title, the issuer, the month, their character as a guide from which organisations choose measures, the intention to supplement them and the content; the summary is written from the Japanese text. The press release, an announcement, gives the day, 27 March 2026, which is in the notes, so date_published is 2026-03; it also confirms the comment period and count and the AI Security Subcommittee. MIC's English press release of the same day, stored on 18 Oct 2026 in run 2026-10-18-recheck-east-asia-1, gives the English title used here and states that the published materials are in Japanese only, so english_version is 'none found'. The annex (別添（付属資料）), published through e-Gov, was not stored. Nothing stored shows a supplement or revision since March 2026.
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 7 field(s) supported.

History

  1. 2026-10-18notes: A draft was open for public comment from 26 Dec 2025 to 29 Jan 2026 and drew 52 → MIC announced the final guidelines in a press release of 27 March 2026; the guid — first sentence added: the day of publication with its source (MIC's press release) and the guidelines' own date; the rest is unchanged2026-10-18-refresh-level-rule-2 · Claude Code (claude-opus-5-5)
  2. 2026-10-18secondary_url: – → https://www.soumu.go.jp/menu_news/s-news/01cyber01_02000001_00282.html — null -> MIC's press release of 27 March 2026, which gives the day of publication, the comment period and count and the AI Security Subcommittee2026-10-18-refresh-level-rule-2 · Claude Code (claude-opus-5-5)
  3. 2026-10-18official_url: https://www.soumu.go.jp/menu_news/s-news/01cyber01_02000001_00282.html → https://www.soumu.go.jp/main_content/001064122.pdf — the record pointed at MIC's press release (報道資料), an announcement; the guidelines' own text, already stored, becomes official_url (rule 7c)2026-10-18-refresh-level-rule-2 · Claude Code (claude-opus-5-5)
  4. 2026-10-18date_published: 2026-03-27 → 2026-03 — 2026-03-27 -> 2026-03: the guidelines' own text gives only the month (cover 令和８年３月); the day, 27 March 2026, comes from MIC's press release, an announcement, and is in the notes (rule 7b)2026-10-18-refresh-level-rule-2 · Claude Code (claude-opus-5-5)
  5. 2026-10-18Re-verified against its sources, no change — pipeline run 2026-10-18-recheck-east-asia-1 (recheck), candidate JP-122026-10-18-recheck-east-asia-1 · Claude Code (claude-opus-5-5)
  6. 2026-10-14Added to the register — found in pipeline run 2026-10-14-discover-leads-2026-10-14, candidate C017 (JP/ministries)2026-10-14-discover-leads-2026-10-14 · Claude Code (claude-opus-5-5)

Related records

AI Guidelines for Business Japan · Published (non-binding, current)

J-AISI Evaluation and Red Teaming Guides Japan · Published (non-binding, current)

Citation

Ministry of Internal Affairs and Communications. Guidelines on Technical Measures to Ensure the Security of AI. 2026-03. https://www.soumu.go.jp/main_content/001064122.pdf. Status: Published (non-binding, current); verified 2026-10-18 (level A). Local copy archive/JP-12/2026-10-18/official.pdf retrieved 2026-10-18 07:42:04 UTC, SHA-256 083731c5a8e96b9a64a08a89f03c435ce620ade2260ca8b8447a59a5bd106d8c. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record JP-12.

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.