Japan's MIC opens comment on a draft election guideline that describes the duty to label AI-made images

Key takeaways

• The comment period runs from 24 October to 24 November 2026, with receipt required by the closing date. MIC's notice asks for comments in Japanese.
• The draft guideline says its listed measures are examples and that operators need not adopt all of them.
• It describes a duty, as the draft puts it, for people who distribute AI-made or AI-altered images or video in election campaign material, or material used against a candidate's election, over the internet, from the official election notice to election day, to ensure a display that the content is AI-made, with two exceptions.
• The draft guideline states that Act No. 58 of 2026 takes effect on 1 March 2027.

What did MIC publish, and how long is the comment period?

On 23 October 2026 MIC announced a public comment process on two drafts: a guideline on appropriate and effective measures to reduce adverse effects on election fairness, and an amendment to the ministerial ordinance that enforces the Information Distribution Platform Act (Ministry Ordinance No. 39 of 2022). The draft guideline is titled 選挙の公正に対する悪影響を軽減するための措置の適切かつ有効な実施に関する指針案.

The comment period runs 令和8年10月24日（土）から令和8年11月24日（火）まで, with receipt required by the closing date, including for postal submissions. MIC's notice says comments must be written in Japanese and can go through e-Gov, email or post. MIC says both drafts are needed to implement Act No. 58 of 2026, which amends the Public Offices Election Act and the Platform Act, and that it plans to prepare the necessary provisions after reviewing the comments.

Is the draft guideline binding, and who would it cover?

Not in draft form. On the draft's own wording, the listed measures are recommended examples, and operators are not obliged to implement all of them. The text says it is issued under Article 27-2(2) of the Platform Act to help large platform operators carry out the mitigation duty in Article 27-2(1). It describes the listed measures as examples and states that they are 全ての措置の実施を義務付けるものではなく, and that specific measures are left to each operator. It also says it is 個別の情報の削除を求めるものではない.

The draft explains that the Diet and the all-party consultation on election campaigning chose this design to protect freedom of expression, so the choice of measures is not subject to administrative disposition or penalties. The Platform Act duties the draft describes are separate. As the draft puts it, the amending Act added provisions requiring designated operators to take measures suited to their service and to publish how they implemented them once a year, and the draft states that the amending Act takes effect on 1 March 2027. We have not read the statute text behind those descriptions.

The draft addresses operators designated under Article 20(1) of the Platform Act. It adds that other services and operators are also expected to take suitable measures.

What does the draft say about AI-made images and deepfakes?

In its background section the draft says it is pointed out that deepfake images made by misusing AI, and the spread of false or mistaken information through foreign influence operations, are undermining election fairness. That is the draft's framing of the problem, not a finding.

Section 4(2) then sets out the labelling duty under Article 142-5 of the Public Offices Election Act as the draft describes it. The draft says that anyone who distributes, over the internet, election-campaign material (or material used against a candidate's election) containing AI-made or AI-altered images or video, where the material is distributed between the official election notice and election day, is subject to a duty to ensure that the recipient's screen correctly shows that the content was made or altered with AI. Whether that duty applies today or only from the amending Act's commencement is not something we resolved. The draft's wording is 正しく表示されるようにしなければならない. Alterations that are trivial by social convention, and content not mistakable for real footage, are excluded. The draft states this duty in the present tense and, separately, that Act No. 58 of 2026 takes effect on 1 March 2027. We did not read the statutes and did not determine which provisions that Act adds or amends.

The same section lists related duties for online posters, such as the ban on publishing false matters about candidates, and criminal provisions on false statements and defamation. The draft presents these as background duties on posters, separate from the platform measures.

What does it recommend platforms do about AI-generated content?

Section 6(4) lists three recommended measures for large operators. First, adopt usage rules on AI-generated content aimed at compliance with the labelling duty, and enforce them. Second, offer or review a function that lets posters mark content as AI-generated on request, or lets users check whether content is AI-generated. Third, detect AI-generated content that the operator considers a risk to election fairness, even where the poster has not labelled it, and show users a warning.

Other recommended measures in section 6 cover staffing for election periods, takedowns and account suspension under terms of service, monetisation limits, showing verified-identity and location information, promoting reliable sources, reducing the ranking of harmful items, and responding to artificial amplification. Section 7 asks operators to publish, preferably with figures, what they did and a self-assessment, in a prescribed format.

What should platforms, campaigns and counsel do now?

This is a consultation, so nothing in the draft is yet a compliance deadline. The practical question is whether to comment and how to prepare.

• Platform operators: compare current AI-content terms, labelling functions and detection practice with section 6(4), and decide whether to file a Japanese-language comment by 24 November 2026.
• Campaigns, parties and political-consulting firms: confirm how AI-made or AI-altered images and video are labelled before distribution during an election period, against the Article 142-5 duty as the draft describes it. Check the statute and its commencement with Japanese counsel.
• Media and AI content providers: identify whether tools you supply can embed or surface an AI marker, since the draft recommends such functions.
• Counsel: obtain the draft ordinance amendment (Attachment 2) and the comment-process notice (Attachment 3) from MIC, and track the final guideline after the comment period.

What we did not verify

What we opened on 25 October 2026 (UTC): the MIC press release page of 23 October 2026, the draft guideline PDF (Attachment 1) and the comment-process notice PDF (Attachment 3). The draft guideline text was extracted from the PDF, and its descriptions of statutes are the draft's own, not our reading of the statutes.

What we did not open: the draft ordinance amendment (Attachment 2, a PDF from which we could extract no readable text), the text of Act No. 58 of 2026, the Public Offices Election Act, the Platform Act, the e-Gov posting, and any English translation by MIC.

What we refuse to claim: that the draft is final, that platforms must adopt any particular measure, that Article 142-5 is new, amended by Act No. 58 of 2026 or already in force, or what the final guideline or ordinance will contain.

Related briefings

• AI Regulation News hub
• Asia-Pacific AI regulation tracker
• AI disclosure and labelling rules
• Japan MIC guideline on deepfake investment advertising
• Brazil TSE election deepfake ban and AI labelling
• Israel Central Elections Committee deepfake marking rules
• India IT Rules on synthetic media labelling