Will Indonesia require AI-generated content to be labelled? The Ministry of Communication and Digital Affairs (Komdigi) says it is studying exactly that. On 17 October 2026, Deputy Minister Nezar Patria said the ministry is examining a ministerial regulation under which AI developers would label the video and other digital media their systems produce, so the public can tell synthetic content from real content. Nothing has been drafted publicly yet, but the direction matters for any organisation that builds AI products, publishes AI-assisted material or relies on video and voice to verify who it is dealing with.
What the deputy minister said
Nezar Patria spoke at the Indonesian Public Relations Convention 2026 (Konvensi Humas Indonesia) in Surakarta on Saturday 17 October, according to CNN Indonesia and Eranasional, both citing Komdigi's official website. CNBC Indonesia and Bloomberg Technoz reported the same remarks on 19 October. He said the ministry is studying a ministerial regulation, a peraturan menteri, so that every AI developer company can provide a label on the authenticity of the video or digital media products it generates.
His reasoning was that the dividing line between fact and synthetic content is becoming thinner. As AI-made audio, images and video become harder to distinguish from the real thing, the risk of fraud and information manipulation rises, CNN Indonesia and Bloomberg Technoz reported. He noted that global technology companies are already developing authentication and content-origin tracing based on metadata, and Eranasional reported that Indonesia is taking a similar approach.
To illustrate the risk, he cited a fraud case in Hong Kong about two years ago in which a company lost US$25 million. According to CNN Indonesia, a finance employee made 15 transfers after an online meeting in which the executives shown were deepfakes. He also said AI should be treated as a tool and a work partner, not a replacement for human reasoning and responsibility.
What is still unknown
Very little detail has been published. None of the reports we read gave a draft text, a timeline, a technical standard for the label, or penalties. The descriptions also differ in firmness. Bloomberg Technoz and CNBC Indonesia describe the rule as under study. CNN Indonesia says it will likely require labels or markers on AI-generated content. Eranasional's headline says Komdigi is finalising the rule and describes developers and platforms embedding authentication markers in visual and audio products, which goes further than the deputy minister's quoted words.
Open questions include who exactly would be covered (only AI developers, or also platforms and organisations that publish AI-generated material), whether text is included, whether the label must be visible, machine-readable or both, and how the rule would relate to the National AI Roadmap. Komdigi is preparing that roadmap for 2026 to 2029 with more than 400 institutions, communities, ministries and industry groups, CNN Indonesia reported.
How the EU and China already label AI content
Indonesia would not be the first. Two large markets already have rules in force, and they show the choices a ministerial regulation would have to make.
- European Union: transparency obligations under the AI Act apply from 2 August 2026, according to the European Commission. Law firm Cooley explains that providers of generative AI systems must embed machine-readable markings and offer a way to detect them, while deployers must disclose deepfakes and AI-generated text on matters of public interest, unless that text has had human editorial review. Marking duties for systems already on the market apply from 2 December 2026, and fines can reach 15 million euros or 3% of worldwide annual turnover.
- The Commission published a voluntary Code of Practice on marking and labelling AI-generated content on 10 June 2026, and Cooley reports that guidelines on the transparency article were adopted on 20 July 2026. The rules reach organisations outside the EU whose AI outputs are used in the EU.
- China: the Measures for Labeling AI-Generated Synthetic Content, issued in March 2025 by the Cyberspace Administration of China and three other agencies, took effect on 1 September 2025, according to law firm Loeb and Loeb. They require visible labels on AI-generated text, images, audio, video and virtual scenes, plus hidden labels in file metadata, with a national standard setting the format.
- In China, app stores must ask whether an app offers generative AI services, and users who publish AI-generated content must declare it and use the platform's labelling functions, Loeb and Loeb reports.
Why it matters for organisations in Indonesia
This section is our analysis. A labelling rule aimed at AI developers would fall first on companies that build generative AI tools. But the questions behind it reach further. Banks, insurers, media groups, retailers and government bodies increasingly generate images, voice and video with AI for marketing, customer service and training. If the Indonesian rule follows the EU or Chinese pattern, it may also ask organisations that publish such content to disclose it, and platforms to preserve or check the labels.
There is also a defensive side. The Hong Kong case the deputy minister cited shows that the risk is not only reputational. Any process that approves payments, opens accounts or changes customer details on the strength of a video call or a voice instruction is exposed to synthetic media, whether or not a labelling rule exists.
What organisations can prepare now
- Map where AI generates content: which tools produce text, images, audio or video, who uses them, and where that output is published or sent to customers.
- Record provenance at the source: keep which model and version produced an output, from what input, and who approved it. That record is the raw material for any label, visible or machine-readable.
- Decide where marking would happen in the publishing pipeline, so that adding a visible notice or metadata later is a configuration change rather than a rebuild.
- Review approval processes that rely on video or voice, such as payment instructions and account changes, and add a second, independent check.
- Follow the consultation: if Komdigi publishes a draft, compare its scope with the EU and Chinese rules, especially if you also serve those markets.
What to watch next
The next signal will be a draft ministerial regulation, which would show who is covered, whether labels must be visible, machine-readable or both, and when the rule would take effect. Watch also for the National AI Roadmap and how it frames AI governance, and for any reference to technical standards for content provenance. In our view, organisations that already know where their AI-generated content comes from will be able to adapt to whichever approach Indonesia chooses with far less disruption than those that do not.
Sources
- Cegah Penipuan, Komdigi Kaji Aturan Wajib Beri Label Konten AI · CNN Indonesia(opens in a new tab)
- Komdigi Susun Aturan Wajib Label Produk AI Biar Warga RI Tidak Ketipu · CNBC Indonesia(opens in a new tab)
- Komdigi Mengkaji Regulasi Label Konten Hasil AI · Bloomberg Technoz(opens in a new tab)
- Komdigi Matangkan Aturan Labelisasi Konten AI demi Tekan Risiko Manipulasi Informasi · Eranasional(opens in a new tab)
- Commission publishes Code of Practice on marking and labelling AI-generated content · European Commission(opens in a new tab)
- EU AI Act: Transparency Obligations Take Effect 2 August 2026 · Cooley(opens in a new tab)
- China's AI Labeling Measures and Mandatory National Standards Take Effect September 1 · Loeb and Loeb(opens in a new tab)
This briefing was prepared by the Indotek editorial desk with AI assistance, from the public sources listed above. It is general information, not legal or professional advice.
AI & Agentic AI
How Indotek can help
Labelling AI content is easiest when a system already knows what it generated, from what and with whose approval. Building that into generative AI applications is the kind of work Indotek does.
- Generative AI on approved content: we build drafting, summarising and translation applications grounded in your own approved content, with a record of the model and inputs behind each output that a label can draw on.
- Humans stay in command: we design AI systems that act inside explicit permissions, log what they do and hand off to a person, with approval points before AI-generated material is published.
- Evaluate before you scale: every AI feature we ship comes with an evaluation set, so output quality, including whether outputs carry the required marking, is measured rather than assumed.
- Workflow automation: we combine deterministic rules with AI judgement in publishing and approval workflows, so exceptions such as unmarked content are handled rather than dropped.
