In short
Indonesia is rewriting its 2014 copyright law, and the draft would make commercial AI training on books and other works subject to a licence. Google, Meta and TikTok pushed back at a parliamentary hearing on 11 November 2026. Sponsors want a plenary vote before the session closes on 12 December.
- Fully AI-generated works would get no copyright; AI-assisted works only if they meet human-involvement criteria
- The draft bans AI imitation of a creator's "distinctive style" and requires disclosure of AI use
- Licences could be signed individually or through a collective management organisation
The hearing was reported by Reuters, which first revealed the AI clauses in July from a draft it reviewed (as carried by The Jakarta Post). On 15 November the trade analyst The New Publishing Standard argued that most English coverage had missed the book-publishing side of the bill. If it passes, Indonesia would be the first country in Southeast Asia to write AI into its copyright law.
The disclosure rule would land on software as much as on writers. Any store, reading app or writing tool that serves Indonesian users with AI features would need a way to record and show where AI was used. Building that into a product is everyday work for teams that handle generative AI integration, and TikTok's objection at the hearing was about exactly where that label should apply.
For novelists, the bill is one more sign that "license first, train later" is becoming a common model outside the United States.
What does the bill say about AI?
The bill would bring AI into copyright law through four rules. According to the draft Reuters reviewed:
- Works made with AI help are protected if they meet human-involvement criteria. Fully machine-made works are not.
- Using AI to imitate a creator's distinctive style is banned.
- AI use in content must be disclosed.
- Copyrighted works may be used for training only under fair use or a licence, and platforms that ignore the rules could lose local business permits.
Hermansyah Siregar, the law ministry official in charge of intellectual property, told Reuters in July that generative AI "has disrupted the copyright framework" and, if unregulated, "could kill human creation." On 27 October, The New Publishing Standard reports, the government added detail: commercial text and data mining would need a reasonable licence, non-commercial research would be exempt, and developers would have to publish summaries of their training data.
Why do Google, Meta and TikTok object?
The three companies objected to different parts of the bill at the 11 November hearing. Reuters reports that Google and Meta focused on rules making platforms pay publishers for news links. TikTok's head of public policy in Indonesia, Noudhy Valdryno, asked that the AI disclosure duty not apply when AI is used only as an assistive tool. He also urged lawmakers to keep safe-harbour protection, saying platforms cannot reliably tell who owns a work or holds its licence.
The New Publishing Standard adds that Google opposed a single collective body issuing blanket licences. Google called licensing all training data unworkable and proposed a text-and-data-mining exception with an opt-out, such as robots.txt.
Why does it matter for books?
The bill goes well beyond news links. According to The New Publishing Standard's reading of Indonesian reporting, the draft covers translation, lending rights and royalties too. It includes a public lending right for authors and more transparent royalty governance. Indonesia is Southeast Asia's largest market, so its choice carries weight in the region.
Governments are picking sides on AI training, and The New Publishing Standard sorts them roughly like this:
- Japan and Singapore: allow AI training under text-and-data-mining exceptions.
- India: has proposed a mandatory licence paid through a central collective.
- Australia: rejected an exception in favour of paid licensing.
- Korea: declined to add an exception.
Indonesia would join the licensing camp. We covered the same split when Australia weighed an opt-out for books and when Korean publishers built a licensed data pipeline.
What does it mean for indie authors?
Indie authors may be best placed under licensing rules, because they usually hold all their own rights. A writer who controls ebook, print and translation rights can opt into a collective scheme, or stay out, without asking a publisher. The open question is how the "human-involvement" test would work. The draft does not say how much human input is enough, or how anyone would check it.
Then there's the style ban. Fan fiction and "in the style of" prompts are common in science fiction, and a law that protects a distinctive style will need careful lines between homage and imitation.
What should authors do now?
- Watch the 12 December deadline: the plenary vote will show whether the licensing and style rules survive.
- Check your distributor terms: if you sell into Indonesia through aggregators, look for new AI or data-mining clauses.
- Keep a rights register: know who owns each format and translation before any collective licence arrives.
- Keep drafts and AI notes: our report on AI disclosure after the Goncourt case explains why working files now matter.



