A modern rule built on an old anchor
Indonesia's House of Representatives (DPR) approved a bill amending the country's Copyright Law — Law No. 28 of 2014 — as a legislative initiative in March 2026, and it has now entered joint deliberation with the government. One clause is doing most of the talking: copyright protection would extend to AI-generated works, but only where there is meaningful human involvement; works produced entirely by machines without human intellectual contribution would remain ineligible. On May 5, 2026, MLex reported that this provision is drawing pointed pushback from legal scholars and creator groups, who warn it could concentrate market power among AI platforms and crowd out human creators.
The worry deserves a fair hearing before it is answered.
The strongest case for the critics
The concentration argument is not frivolous. If AI-assisted output becomes copyrightable on a thin showing of "human involvement," the entities best placed to mass-produce and claim ownership over that output are large platforms and well-capitalised studios that run generative models at scale. They can flood the markets for music, stock imagery, and text with protectable AI-assisted works, depressing prices and demand for purely human craft — often using models trained on the very creators now competing against them, with no remuneration flowing back. Worse, when the legal standard for "human touch" is left undefined, the party with the deeper legal bench tends to win the ownership fight. That asymmetry favours platforms over individual artists. As a matter of distributive impact, the critics have identified something real.
Why the human-authorship anchor is still the right call
The instinct behind the bill, however, is correct — and abandoning it would make the very harm critics fear worse. Under the existing Law No. 28 of 2014, copyright vests automatically in an "author," defined as "a person or several persons who individually or jointly produce works that are unique and personal" (WIPO Lex). Indonesian practitioners broadly agree that this human-authorship requirement already excludes purely machine-generated output from protection. The bill does not invent a new entitlement so much as draw a boundary the statute always implied: human creativity in, pure automation out.
That boundary is precisely what prevents pure machine output from being monopolised. Academic proposals went considerably further — a 2024 paper in the journal Law Reform urged Indonesia to grant AI a "limited artificial legal persona" so that rights could attach to the developers or operators of a system. The bill wisely declined that path. Granting copyright to machine output as such — assignable to whoever owns the model — is the scenario that would genuinely concentrate power. Requiring a human author is the firewall against it, not the cause of it.
Denying protection to all AI-assisted work, the alternative some critics imply, would also misfire. It would strip legitimate human creators who use AI as a tool — the photographer who composites, the musician who arranges, the designer who iterates — of any protection, and push AI use into undisclosed grey zones. A blanket ban on AI-touched works punishes the human in the loop to spite the machine.
The real fault line is the undefined threshold
The genuine defect is not the rule; it is its vagueness. As Indonesian counsel have noted, the concept of "human involvement" is central to the bill but "currently lacks a detailed legal standard" — there is no articulated threshold for how much human intervention qualifies (Bagus Enrico & Partners). A one-word prompt and a thousand-word edit would, on the statute's face, sit on the same side of an invisible line. That indeterminacy is exactly what hands the advantage to well-resourced litigants and exactly what the concentration critics are reacting to. The problem is curable with precision, not with retreat from the principle.
Proportionate regulation would do four things. First, define the threshold objectively — borrowing from the US Copyright Office's 2025 guidance, which protects the human-authored selection, arrangement, and modification of AI output rather than the raw generation. Second, require disclosure: works claiming protection should declare material AI assistance, giving courts and competitors a factual basis to test "human involvement" rather than a lawyering contest. Third, address the input side the output rule ignores — the training-data question of whether scraping creators' works is fair use or requires licensing is where creator livelihoods are actually decided, and the bill is largely silent on it. Fourth, leave concentration to competition law, which is built to police market power; copyright scope is a clumsy and over-broad instrument for industrial policy.
A broader modernisation worth getting right
The AI clause sits within a wider, overdue update. The same bill strengthens collective royalty management — responding to the Constitutional Court's December 2025 ruling in Decision No. 28/PUU-XXIII/2025, which clarified that commercial event organisers, not performers, owe performance royalties — and extends platform liability to user-generated-content services while preserving fair-use carveouts for education, research, parody, news reporting, and disability access (Asia IP). The Directorate General of Intellectual Property has framed the reform as a push for "transparent, accountable" governance that adapts to technology while protecting the creative economy (DGIP).
That is the right ambition. Indonesia should keep the human-authorship anchor, which is pro-creator and pro-innovation at once, and spend its remaining deliberation defining what "human involvement" means — clearly enough that the answer does not depend on who can afford the better lawyer.