AI & Automation

Indonesia's Two AI Regulations Await Signature: Sort Your AI Uses by Risk Now

November 13, 2026 · 6 min read · Etzal Group Team

Two draft presidential regulations on artificial intelligence in Indonesia are said to be waiting only for the President's signature. Edwin Hidayat Abdullah, Director General of Digital Ecosystems at the Ministry of Communication and Digital Affairs (Komdigi), said so at an OpenAI event on Thursday, 3 September 2026, as reported by the technology news site Selular.ID the next day. As of mid November 2026, when this article was written, the sources we read do not report either one as issued.

For operations, IT and finance leaders, that means the direction of Indonesia AI regulation is visible while the wording is not. That is enough to start work, provided you do not invent the content of rules that do not yet exist.

What has been stated

A Peraturan Presiden, or Perpres, is a regulation issued by the President. According to Selular.ID there are two drafts:

  • A Perpres on the White Paper and Roadmap for Artificial Intelligence, covering 2026 to 2029.
  • A Perpres on Artificial Intelligence Ethics.

Edwin Hidayat Abdullah is quoted as saying, in Indonesian, that only the President's signature remains. On the target of finishing both this year, he said that was the ministry's hope.

On content, Selular.ID reports that the ethics regulation will set limits on AI use in three groups: prohibited uses, uses that need strict oversight, and lower-risk uses. More technical rules will come from the relevant ministries and agencies in their own fields. Industry and academics took part through public consultation, and industry players are said to have received the draft.

There is one more recent item. Katadata reported on 13 November 2026 that the government is preparing a cross-sector national AI roadmap as part of its digital transformation agenda, according to remarks by Komdigi Secretary General Ismail at an ITU conference in Doha on 11 November 2026.

What is not known

This is where care is needed.

  • Both regulations are still drafts. The timing is an official's hope, not a binding schedule.
  • The source does not say which uses fall into the prohibited group, which need strict oversight, or what the tests are. Lists of examples circulating outside official sources are best treated as guesses.
  • Concrete duties for companies, transition periods and penalties are not known from the sources we read.
  • Sector rules will follow from ministries and agencies, so companies in heavily regulated sectors will probably face more than one rule.

Because the status can change at any time, check the latest official announcement from Komdigi and the State Secretariat before acting on this article.

One obligation with a date already attached

While the AI regulations wait, another rule reaches AI through personal data. According to a summary published by RSM Indonesia on 18 September 2026, the elucidation of Government Regulation Number 33 of 2026 names AI and machine learning as new technologies that trigger a data protection impact assessment (DPIA), from 16 January 2027.

Note that this comes from a consultancy's summary, not from an official text we read ourselves. Read the official text or ask your legal adviser to confirm its wording and scope.

The practical point: if your AI processes personal data about customers or employees, the inventory described below may be needed sooner than any AI regulation.

Start with an inventory

This section is our advice. No rule prescribes the format below.

Most mid-sized companies do not know exactly where AI is used. There are AI features in software already bought, staff using AI assistants on personal accounts, and pilots that were never closed. Build one table with these columns for each use:

  • Name of the use and the unit that owns it.
  • Tool or vendor.
  • Data that goes in, and whether any of it is personal data.
  • Output produced, and who reads it.
  • Whether the output becomes a decision about a person, or only material for a human who decides.
  • Who checks the results, and how often.

Group them provisionally by effect

We do not yet know the tests the regulation will use. But the three reported groups show the direction: the greater the consequence of a use for people, the tighter the oversight. On that basis you can keep a provisional internal grouping:

Internal groupCharacteristicsAction now
Needs leadership reviewProduces decisions about people without human checking, or uses sensitive personal dataPause expansion, request legal review
Needs oversightAffects customers, employees or money, with a human checkingName an owner, record the checks, keep logs
Lower riskSupports internal work, output always edited by a personRecord in the inventory, review yearly

This grouping is our internal working tool, not the content of the regulation. When the rules are issued, you remap existing rows to the official groups. That is far lighter than starting from nothing.

Implications for the 2027 AI budget

Three decisions are worth making before the budget closes.

Hold expansion of the uses that affect people most. Not stopped, only not widened until the wording is clear. Lower-risk uses need not wait.

Ask for an adjustment clause in contracts. For multi-year AI purchases, ask the vendor to commit to adapting its product to Indonesian rules issued later, and agree who bears the cost.

Budget for record keeping. Logs of what went in, what came out and who approved are material that almost any form of oversight will ask for. Building them in from the start costs less than adding them later.

The trade-off: holding expansion delays the benefit. For some companies the risk of rework is smaller than the cost of waiting. That is a legitimate business decision, as long as it is made knowingly and the reasons are recorded.

What you can do this week

  1. Name one person as owner of the AI inventory.
  2. Send a short form to every unit head: which AI tools are used, for what, and with which data.
  3. Ask your main software vendors for a list of AI features in their products.
  4. Sort the results into the three internal groups above.
  5. Schedule a session with legal counsel on the DPIA obligation and to revisit the inventory when the regulations are issued.

This article is not legal advice. Confirm your company's case with a legal consultant or the relevant authority.

Etzal Group helps companies build inventories and governance of this kind through its AI consulting service.

Sources

Frequently asked questions

Are Indonesia's AI regulations in force?

Not according to the sources we read. On 3 September 2026 the ministry said both drafts only await the President's signature and it hopes they will be finished this year, so check the latest official announcement.

What are the three groups of AI use in the draft ethics regulation?

Selular.ID reports that the draft will distinguish prohibited uses, uses that need strict oversight, and lower-risk uses. The detail of each group was not published in that source.

What can a company prepare before the rules are issued?

List every use of AI, record the data and decisions each one touches, and group them provisionally by their effect on people. That list stays useful whatever the final wording says.

Author

Etzal Group Team

Etzal Group Team, helping businesses grow with technology. Contact us.

Related service

WhatsApp AI Chatbot

We build, train, and run a WhatsApp AI agent that answers customers 24/7, qualifies leads, and follows up.