Part of the AI Regulation News hub.
Seven Japanese government agencies have jointly asked Google, LINE Yahoo, Meta, TikTok Japan and X to verify advertiser identity, label AI-generated ads and report their numbers by fixed dates
The document tells you in a footnote that it binds nobody. It then sets three reporting deadlines, names the reporting inbox, and specifies the metrics down to average response time.
Bottom line: Not binding. Footnote 2 of the request states that it constitutes administrative guidance under Article 2(6) of the Administrative Procedure Act (Act No. 88 of 1993) and does not constitute a disposition under Article 2(2). The asks are framed as voluntary measures.
Who this affects: Trust and safety leads, ad policy owners and Japan country counsel at Google LLC, LINE Yahoo Corporation, Meta Platforms, TikTok Japan and X Corp. Compliance officers at registered financial instruments business operators have an indirect interest, because the request cites the Financial Instruments and Exchange Act.
Issue date: 7 August 2026. Written descriptions of the measures are requested by 16 October 2026. The measurement window runs 1 October 2026 to 28 February 2027, and the results report is requested by 16 March 2027.
What changed: The named platforms are asked to verify advertiser identity using electronic authentication, to surface advertiser and ad information including whether an ad was produced mainly with generative AI, to act on takedown requests without delay, and to report specific counts and ratios to the Digital Agency.
Analysis: The reporting schedule is the instrument. A non-binding request that asks for deletion counts split by legal ground and by trigger, plus the ratio of deletions to requests per requesting body and average handling time, produces a dataset that a future binding rule can be calibrated against. Voluntary now, measured now, is how the baseline gets set.
Primary sources: MIC press release (JA) · Text of the joint request, all five addressees (JA, PDF)
- Instrument (EN)
- Request concerning the strengthening of measures against impostor scam advertising on SNS and other interactive platforms (working translation)
- Authority
- National Police Agency, Financial Services Agency, Consumer Affairs Agency, Digital Agency, Ministry of Internal Affairs and Communications, Ministry of Justice, Ministry of Economy, Trade and Industry. The Digital Agency is the coordinating body
- Jurisdiction
- Japan
- Status
- Issued and delivered in writing to five named companies
- Bindingness
- Administrative guidance under Article 2(6) of the Administrative Procedure Act; expressly not a disposition
- Issue date / next deadline
- Issued 7 August 2026; descriptions of measures due 16 October 2026; results report due 16 March 2027
- Addressees
- Google LLC, LINE Yahoo Corporation, Meta Platforms Inc., TikTok Japan, X Corp.
- Reporting channel
- Digital Agency Strategy and Organisation Group, in writing
- Primary source
- https://www.soumu.go.jp/menu_news/s-news/01ryutsu02_02000486.html
The three asks
Item 1 is advertiser identity verification. Platforms are asked to reliably verify advertisers, including through electronic authentication means, so that advertisers can be identified and traced. A footnote names the My Number Card as one reliable means for individuals and the commercial registration electronic certificate as one reliable means for corporations.
Item 2 is ad transparency. Consumers should be able to check information about the advertiser, meaning name, location and whether identity verification occurred, and information about the ad itself, meaning that it is an ad, that it was produced mainly using generative AI technology, and why it is being shown.
Item 3 is takedown discipline. Where an impostor scam ad may violate the Penal Code, the Financial Instruments and Exchange Act or other laws, and a removal request arrives from the Internet Hotline Center or a competent ministry, the platform is asked to make a decision without delay and act promptly and appropriately.
Two legal hooks the request is careful to name
Footnote 2 sets out the criminal theory. Where a person displays fabricated content such as a celebrity's name used without permission alongside false investment results, and circulates or uses that forged electromagnetic record relating to proof of fact online, it may violate Article 161(1) of the Penal Code, the offence of using forged private electromagnetic records.
Footnote 3 sets out the securities theory. Where an unregistered operator uses an ad that on its face does not itself amount to financial instruments business as an entry point, and then conducts acts amounting to financial instruments business on the website or platform to which viewers are led, the whole sequence may violate Article 29 of the Financial Instruments and Exchange Act.
Neither footnote asserts that any specific ad broke either law. They tell platforms which statutes their reviewers should be reasoning from, which is a different and more useful thing.
The number that explains why these five companies
A footnote cites National Police Agency finalised figures for 2025 on recognised cases of special fraud and SNS-type investment and romance fraud. Among initial contact routes for SNS-type investment fraud, the shares attributable to banner and similar advertising were YouTube 27.1 percent, Instagram 16.3 percent, TikTok Japan 10.4 percent, Facebook 6.5 percent, LINE 4.8 percent and X 2.1 percent.
Add those and roughly two-thirds of the measured advertising entry points sit with the five addressees. The recipient list is not a political gesture, it is the distribution.
What the platforms are being asked to count
For identity verification: implementation counts and ratios by verification means, split domestic and overseas, plus the count and ratio of listings rejected for defective verification, by means.
For transparency: what information was actually shown on ads, and the count and ratio of ads by the type of information shown.
For takedowns: deletion counts by reason, meaning Penal Code violation, Financial Instruments and Exchange Act violation or other, and by trigger, meaning a request from the Internet Hotline Center or a competent ministry, a voluntary deletion, or a report from an ordinary user. Where the trigger was a removal request, platforms are also asked to report, per requesting body, the ratio of deletions to requests, average handling time, and a breakdown by reason of the cases not deleted.
That last clause is the one to read twice. Reporting why you declined is a heavier obligation than reporting what you removed, and this request asks for it voluntarily.
The two guideline revisions this request sits on top of
Footnote 2 also records a sequencing detail worth noting. The Internet Hotline Center's Hotline Operation Guideline was revised on 7 August 2026, and MIC's guideline on Article 26 of the law dealing with rights infringement arising from information distribution via specified telecommunications was revised on 6 August 2026, in each case to newly specify this conduct as illegal information.
So the guideline changes landed first, on 6 and 7 August, and the joint request followed the same day as the second one. The request is asking platforms to act quickly on removal requests that the revised guidelines have just made easier to send.
What we did not verify
I opened the MIC press release of 7 August 2026 and the full 15-page annex PDF containing the request as addressed to Google LLC, LINE Yahoo, Meta Platforms, TikTok Japan and X Corp. All dates, deadlines, statutory citations, percentages and reporting metrics above come from that PDF.
I did not open the National Police Agency statistical release the percentages are drawn from, the revised Internet Hotline Center operation guideline, the Penal Code or Financial Instruments and Exchange Act provisions cited, or any platform response. No platform had published a response that I looked for.
I will not claim that any of the five companies has agreed to any of this, that any ad has been found unlawful, or that non-compliance carries a sanction. The request says on its face that it is not a disposition, and I have seen no enforcement instrument attached to it.
If you own ad policy for a large platform operating in Japan, calendar 16 October 2026 and 16 March 2027 now, because the measurement window opens 1 October. The generative AI labelling ask is the sleeper: item 2 wants users to be able to see that an ad was produced mainly with generative AI, which is a product change, not a policy change. And a request that is voluntary today is generating the exact statistics a binding rule would need tomorrow.
Source File
https://www.soumu.go.jp/menu_news/s-news/01ryutsu02_02000486.html
Open the annex PDF linked above and confirm four things: the date line reads Reiwa 8, 7 August 2026; the footnote marked with a double asterisk on the second page of each letter says the request is administrative guidance under Article 2, item 6 of the Administrative Procedure Act and not a disposition; item 2(1) asks that ads disclose that they were created mainly using generative AI; and item 3(3) sets 16 March 2027 for the deletion statistics.
This request constitutes administrative guidance as provided in Article 2, item 6 of the Administrative Procedure Act (Act No. 88 of 1993), and does not constitute a disposition (item 2 of the same Article). Joint request of the NPA, FSA, CAA, Digital Agency, MIC, MOJ and METI, 7 August 2026, footnote 2, author's translation from the Japanese
FAQ
Can a platform be penalised for ignoring this request?
Nothing in the document provides for that. It states that it is administrative guidance and not a disposition under the Administrative Procedure Act, and the asks are framed as strengthening of voluntary measures.
Does it require platforms to label AI-generated advertising?
It requests, not requires. Item 2 asks platforms to take measures so that users can check ad information including that the ad was produced mainly using generative AI technology, alongside advertiser name, location and whether identity was verified.
Which laws does the request say impostor scam ads can breach?
Two are named. Article 161(1) of the Penal Code, on use of forged private electromagnetic records, where a celebrity's name is used without permission with fabricated investment results. And Article 29 of the Financial Instruments and Exchange Act, where an unregistered operator uses an ad as an entry point to conduct financial instruments business elsewhere.
Why were these five companies chosen?
The request cites National Police Agency 2025 figures on the advertising contact routes for SNS-type investment fraud: YouTube 27.1 percent, Instagram 16.3 percent, TikTok Japan 10.4 percent, Facebook 6.5 percent, LINE 4.8 percent and X 2.1 percent.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.