Corrections
We do not delete an error. When a figure turns out to be wrong the piece keeps its address, the text is corrected, and the change is written down here. What follows is the whole record, not a selection from it.
Every correction, newest first
Entries: 5The first version said Law 7 of 2021 could not be retrieved, graded the reading of Article 6(1) an estimate for that reason, and recorded Article 32A as unread. The law was obtained on 20.10.2026 from the tax authority's own file host. Article 6(1) keeps the words that confine the deduction to a domestic taxpayer and a permanent establishment, so that claim is now verified. Article 32A was read and turns out to be a power to conclude tax agreements, stating no rule that gives one priority over Article 26.
WhyA sibling piece published six days later reached the same law from a different host, which put the registry and this article in contradiction one click apart. Our failure was a fact about one endpoint on one day, and the article stated it as a fact about the document. The two provisions it hedged on are both amended by that law, so the hedges could not be left standing once the text was in hand.
The newsroom
The byline read Made Rai. It now reads Komang Wirata, of The Map desk. The text was not changed.
WhyThe desks this publication lists against an author are a statement about which subjects that byline is competent on. Made Rai is listed for compliance and title, not for The Map, and the piece was published under that byline anyway. A gate now refuses an article whose desk is not one its author covers.
The newsroom
The byline read Elena Hartono. It now reads Bayu Prasetyo, of the Build desk. The text was not changed.
WhyElena Hartono is listed for occupancy and The Index. This is a reading of building-permit law, and at publication no author was listed for the Build desk at all — which is how the mismatch happened rather than being caught.
The newsroom
The byline read Kenji Tanaka. It now reads Rina Halim, of the Resident desk. The text was not changed.
WhyKenji Tanaka is listed for compliance and exit. This piece is about what a stay permit permits, and the Resident desk had no author listed when it was published.
The newsroom
The first version gave the gross yield as 5.80 per cent, from first-quarter data. It was replaced with 4.24 per cent, from the August 2026 survey.
WhyFresher data from the same source had been published. The old figure would still have been correct, but out of date.
The newsroom
A correction is a record that the standard failed. The standard itself, and what to do when a figure looks wrong to you, is set out in how we check facts.