Depublishing

Depublishing means: a text, video, or audio piece is put online and later deleted again. The word is the opposite of publishing, i.e., making something public. What matters is the passage of time. The content doesn’t disappear because it was wrong or violates a law. It disappears because a previously set deadline has expired. The term became well known in Germany through the public broadcasters ARD and ZDF, which for years had to remove their online content from their media libraries after a fixed period.

The dispute over broadcasting fees and newspaper publishers

Behind depublishing lies an economic conflict. ARD and ZDF are financed through the broadcasting fee, which almost every household in Germany pays. Newspaper publishers, on the other hand, finance themselves through subscriptions and advertising. The publishers argued: if fee-financed broadcasters put unlimited free texts online, no publisher can make money from journalism anymore.

The legislature responded in 2009 with the so-called Twelfth Interstate Broadcasting Amendment Treaty. It set fixed retention periods for online content. Many pieces were only allowed to stay online for seven days, others for six or twelve months. After that, they had to be deleted. Critics called this a deliberate destruction of knowledge that the public had already paid for.

Since 2019, the rules have been significantly relaxed. Most content may now remain online permanently. Nevertheless, depublishing still occurs, for example with purchased films and sports rights. These contents are subject to licensing agreements that only permit a certain period of use.

Retention periods, licenses, and dead links

Technically, depublishing is simple. Every piece of content is given an expiration date upon publication, the retention period. An automated system regularly checks which deadlines have expired and removes the affected pages. Sometimes the page is deleted entirely, sometimes a notice remains in its place.

For users, this creates dead links. Anyone who linked to a TV segment in an old forum post or in a term paper will later find nothing there. Experts speak of link rot, i.e., the decay of references. Archives such as the Wayback Machine of the Internet Archive therefore store copies of web pages so that old content remains findable.

Depublishing should not be confused with the right to be forgotten. In that case, an individual person demands that information about them disappear from search results. Depublishing, by contrast, concerns entire pieces of content regardless of who appears in them. Deletion due to an error is also something different: that is called a correction or retraction.

From media libraries to AI training datasets

In everyday life, the phenomenon is mostly encountered in media libraries. A series remains available until a certain date, then it’s gone. Streaming services work similarly, removing films from their catalog once the license expires. Authorities and companies also depublish, for example old press releases or expired job postings.

In tech news, the term increasingly appears in connection with artificial intelligence. Language models are trained on huge amounts of web text. If a text is later depublished, its content still remains embedded in the model. Knowledge is very difficult to remove from a trained model once it’s in there. Researchers call this attempt machine unlearning.

For the archiving of history, depublishing is a real problem. Historians can look up newspapers from 1950 in a library. A news site from 2015, on the other hand, has often vanished irretrievably. That’s why libraries and scholars have for years been calling for legally mandated web archives for Germany.