Artificial Intelligence Policy

The Bali Observer uses artificial intelligence (AI) in producing its daily news copy. We say so here, once and for the whole site, because AI assistance is part of how we work every day, not an exception that applies to particular articles. This page explains what AI does, what the newsroom does, and what we will not do. Responsibility for every story stays with people: the reporter in the byline, the Responsible Editor, and the press company that publishes this site.

Basis of this policy

This policy refers to Press Council Regulation Number 1/PERATURAN-DP/I/2025 on Guidelines for the Use of Artificial Intelligence in Journalistic Work, adopted on 22 January 2025. Its main provisions:

• journalistic work produced with AI remains bound by the Journalistic Code of Ethics;
• there must be human control from start to finish;
• the press company is responsible for the work;
• data, information, images, audio, and video obtained through AI must be checked for accuracy and verified;
• any use of AI that significantly affects the work must be clearly disclosed;
• manipulated images, avatars, and AI-generated voices must be labelled, and imitating a specific figure requires the consent of that person or their heirs;
• copyright and privacy must be respected;
• disputes are settled through Press Council mechanisms under the Press Law.

This policy is also subject to Law Number 40 of 1999 on the Press, the Cyber Media Reporting Guidelines, and Law Number 27 of 2022 on Personal Data Protection.

What we use AI for

• Monitoring press releases, institutional documents and data, and publicly available coverage to find events in the areas we cover.
• Drafting news copy from those open sources. Every source used is named and linked in the "Sources" block of the article.
• Cross-checking the content of each story against at least two independent sources before it is published.
• Tidying language: spelling, sentence structure, consistency of terms, and suggestions for headlines, summaries, and tags.
• Support tasks such as transcribing recordings, translating foreign-language material, and processing data tables.

Workflow: from source to edit

1. The newsroom sets the rules. Coverage area, sections, which sources may and may not be used, the verification threshold, and house style are decided by people. The system applies those rules; it does not make them.
2. Open-source material. Copy is drafted only from sources a reader can open. It is written in our own structure and sentences, never copied from the source.
3. Two-source verification. Names, titles, figures, dates, and quotes must be found in the source. A story not supported by at least two independent sources is not published and is replaced by another. Drafting and cross-checking at this stage run with AI assistance under the newsroom's rules.
4. Publication. Copy that passes verification is published with the byline of the reporter who holds that section's desk, a Sources block, and the actual time of publication.
5. Edited by the newsroom after publication. We publish first, then human editors read and edit the published copy: fixing it, updating it, or withdrawing it where necessary. Substantive changes are marked "Updated" with the time.
6. Open correction. Errors found by the newsroom or by readers are fixed in the open with a "Correction" note, under our Corrections Policy.

We describe this sequence as it is: human editing happens after a story is published, not before. That is why two-source verification is a condition of publication, and why we handle corrections without delay.

What we will not do

• Publish a story that has not passed the cross-check against two independent sources.
• Treat an AI answer as a source. AI can be wrong and can invent. Sources remain documents, data, and people.
• Invent quotes, sources, reporter names, data, or events.
• Use AI-generated images, video, or audio to depict real events, people, or places.
• Imitate anyone's face or voice, including as an avatar or dubbing, without the consent of that person or their heirs.
• Put the identity of a confidential source, an embargoed document, or unnecessary personal data into a third-party AI service.
• Hand AI the newsroom's editorial stance or the judgement of a person.
• Credit AI as an author.

Disclosure to readers

We do not attach an AI note to each article. Our disclosure is made at site level: this page applies to all news copy at baliobserver.com and is linked from the footer of every page. What every article does carry is what readers can check for themselves:

• a "Sources" block linking to the sources used;
• the byline of the reporter responsible for that story;
• the actual time of publication, and an "Updated" or "Correction" mark when the content changes.

Should we ever publish a conceptual AI illustration — for an abstract topic that cannot be photographed — it is labelled "Illustration generated with artificial intelligence" directly beneath it and will not resemble a real person or event.

The "Listen" feature

The "Listen" button on articles uses the text-to-speech engine built into your browser or device. The voice you hear is your device's synthetic voice, not a recording of a journalist or a source, and it imitates no one.

Bylines and responsibility

The name in a byline is the newsroom reporter who holds that section's desk and is responsible for the story, including editing and correcting it after publication. A byline never names a tool. The team is listed on our Masthead page, and the full byline rules are in Editorial Standards. Institutionally, PT. Kabar Delapan Belas as the press company, and the Responsible Editor, Mursyid Sonsang, are responsible for everything published at baliobserver.com, including work produced with AI assistance. Errors cannot be shifted onto a technology provider.

Copyright, privacy, and security

• AI output is handled carefully so as not to infringe copyright. Articles are written in our own words, sources are named and linked, and other people's work is not copied.
• Personal data is processed under our Privacy Policy. AI is not used to profile readers. The Bali Observer currently runs neither content personalisation nor programmatic advertising.
• The newsroom selects AI services that are secure and reliable, and reviews their use periodically.
• Should advertising material produced with AI ever appear, it will be labelled.

Corrections, complaints, and disputes

If you find an error in an article, a source that does not support the story, or an AI-generated image or voice that is not labelled, write to redaksi@baliobserver.com. Complaints are handled under our Corrections, Right of Reply & Right of Correction Policy. Disputes over journalistic work involving AI are settled through the Indonesian Press Council.

Review

This technology changes fast. The newsroom reviews this policy at least once a year, and sooner if the Press Council issues new rules. Our wider working standards are in Editorial Standards & Fact Verification.

This page is reviewed regularly by the The Bali Observer newsroom. Responsible Editor: Mursyid Sonsang.