AI and Copyright in Indonesia 2026: What Creators and Brands Should Prepare as New Rules Are Debated

A designer uses AI to explore dozens of visual directions before redrawing a concept.

A copywriter asks AI for five structural alternatives, then rewrites the entire piece.

A brand generates campaign imagery with generative AI.

An agency places client references into an AI service to develop ideas.

All four workflows involve AI.

Their copyright, contractual, confidentiality and creative-provenance risks are not the same.

The distinction matters increasingly in Indonesia. Law No. 28/2014 on Copyright remains in force, while Parliament has advanced a proposed amendment as a DPR initiative in 2026. Generative AI and the use of creative works as training data are among the issues policymakers are examining.[1][3][4]

At the end of July, Komdigi explicitly identified the use of creative works for AI training as one of the difficult issues that needs to be balanced against technology innovation and creators' economic rights.[5]

For creators and businesses, regulatory uncertainty is not a reason to stop using technology.

It is a reason to improve creative governance now.

Separate current law from rules still under debate

The first distinction matters.

Law No. 28/2014 remains in force.[1] Proposed amendments and policy discussions should not be treated as final legal requirements.

Businesses should therefore avoid treating individual proposals from the copyright debate as though they are already enforceable law.

The debate is still useful as a signal of where risk is receiving greater attention:

training data, human contribution, economic rights, digital platforms, AI-enabled works, attribution and provenance.

Specific disputes or material contractual decisions should still be reviewed with qualified intellectual-property counsel.

AI-assisted and AI-generated work are not the same question

DJKI said in April that, within the legal framework under discussion, AI is being treated as a tool rather than a legal subject. One important unresolved issue is the threshold of human involvement required for a work to qualify for copyright protection.[6]

A DJKI official has also stated that Indonesia's existing law continues to place humans at the centre of authorship and that work produced entirely by a machine without human creative involvement is not treated as a protected work under that interpretation.[7]

For a creative team, the practical response is not simply to attach the label “made with AI”.

A better question is:

What did the human actually contribute to the final work?

Did a person develop the concept?

Select and arrange elements?

Redraw or materially edit the output?

Rewrite it?

Direct the composition?

Exercise meaningful creative selection?

Produce the final artwork?

The more complex the workflow, the more useful documentation becomes.

Keep evidence of the creative process

Copyright protection in Indonesia generally arises automatically when an eligible work is expressed in tangible form. DJKI nevertheless encourages copyright recordation because it can provide stronger evidence of ownership if a dispute occurs.[2]

Recordation is not the only useful evidence for a business.

Keep:

creative briefs, rough sketches, source files, version history, revision dates, approval emails, working files and contracts.

For AI-assisted workflows, consider also recording:

which tool was used, where AI entered the process, what it produced and what material creative changes were subsequently made by people.

The objective is not to create excessive creative bureaucracy.

It is to make the provenance of a work explainable when it matters.

Agencies should establish ownership before the project begins

Copyright ownership questions existed before generative AI.

AI simply makes some of them more visible.

Who owns material produced by a freelancer?

What happens when an agency uses licensed stock assets?

Does the client receive all intellectual-property rights or only specified usage rights?

May the agency display the work in a portfolio?

Who is responsible for licensing fonts, photography, music, footage or illustration?

Generative AI adds another layer:

Can confidential client material be entered into an external AI service?

Do the platform's terms permit the intended commercial use?

Does a final asset contain elements that require further clearance?

Contracts should allocate responsibilities before a dispute arises.

The appropriate wording will depend on the parties and use case; this article does not prescribe legal clauses.

Do not feed confidential assets into AI simply because it is convenient

Copyright is only one part of the risk.

Creative teams often hold unreleased products, campaign strategies, customer information, source files, talent photography, brand plans, pitch decks and confidential client material.

Before entering any of those into an external AI platform, the company should understand both its own internal rules and the relevant service terms.

The basic question is:

Do we actually have the right and permission to put this information into that system?

When the answer is unclear, the convenience of one prompt may not justify the exposure.

Brands must also consider similarity and trademarks

Brands using AI face more than copyright questions.

AI can produce names, logos and visual identities at enormous speed. The ability to generate a sign, however, does not itself create exclusive trademark rights.

DJKI has reiterated that Indonesia follows a first-to-file trademark system: exclusive rights arise through registration rather than simply because a name or logo was created or used.[8]

When AI produces a candidate name or logo, therefore, do not assume the result is clear for commercial use.

Appropriate trademark search and clearance remain important before substantial investment in a new identity.

The ability to generate 1,000 logos in an hour also increases the importance of checking whether a result is uncomfortably similar to something already protected.

Training data is an area to watch closely

Komdigi has explicitly identified the use of creative material as AI training data as one of the major copyright-policy challenges being considered.[5] Members of Parliament have likewise raised concerns about the use of works for training without authorisation.[4]

The article should not predict Indonesia's final answer.

Questions such as whether licences are always required, which exceptions may apply, how opt-out mechanisms could work, how remuneration might be structured and how cross-border models should be treated remain part of an evolving policy landscape.

Creators and publishers can nevertheless improve their records of what they own and control.

Companies building or fine-tuning AI systems should pay increasing attention to dataset provenance and appropriate legal review.

Human review should become more systematic, not disappear

AI can produce 100 ideas in the time a creative team once produced five.

Volume is not quality assurance.

Brands still need to assess:

factual accuracy, originality concerns, visual similarity, trademarks, cultural sensitivity, representation, licensed elements and confidential material.

The easier content becomes to generate, the greater the temptation to reduce review because production volumes increase.

That is the wrong direction.

Faster creation requires more systematic quality control, not less.

Give creative teams an internal AI use policy

A policy does not have to be a 40-page manual.

A small agency can start with one page.

For example:

Green — generally permitted

Brainstorming, outlines, internal ideation and proofreading of non-confidential material.

Amber — human review or approval required

Client-facing copy, visual exploration, campaign concepts, research summaries and code.

Red — do not enter without specific permission

Confidential client materials, personal data, unreleased assets, passwords, credentials and proprietary datasets.

These categories are an operational example, not legal rules.

The objective is to prevent every employee from inventing an individual standard.

AI documentation also protects a brand from its own memory

Six months after a campaign launches, the person who produced it may have moved to another team.

A client then asks:

Where did this photograph come from?

Who licensed the music?

Which part was generated with AI?

Who created the final artwork?

If the answer exists only in one designer's memory, the company already has an operational risk.

A simple asset provenance record can capture:

asset name, creator or source, licence, AI involvement, human editor, approval, permitted uses and source-file location.

For a small company, a spreadsheet may be sufficient.

What matters is the existence of a trail.

Do not wait for final regulation to adopt practices that already make sense

Indonesia's policy framework is still evolving.

The copyright law amendment is being debated, while the government is also developing broader national AI governance.[3][5]

But many of the practices businesses need are not merely defensive regulatory measures.

They are sound operational practices:

know who created an asset;

know who owns the rights;

know which tools were used;

protect confidential material;

verify licences;

document meaningful changes;

and ensure that a human remains accountable for the final result.

AI does not reduce the value of human creativity—it increases the need to explain it

When software can generate text, images, music and video in seconds, creative value is no longer only about producing an output.

More value moves into:

judgment.

Taste.

Direction.

Context.

Original contribution.

And responsibility for the final work.

For creators, agencies and brands, the challenge in 2026 is not to choose between “AI” and “no AI”.

It is to build workflows in which technology expands human capability without eliminating clarity about provenance, ownership and accountability.

Regulation will continue to evolve.

Businesses that establish that clarity now will be in a stronger position than those that begin reconstructing their source files only after a dispute has already arrived.