ZDF: Report with fake AI video in "heute journal"
On Sunday, "heute journal" illustrated a report on ICE operations with a video generated by Sora. The broadcaster initially explains it as intentional.
In the corrected version of the report, ZDF has removed the AI material and refers to a change for editorial reasons.
(Image: Screenshot/ZDF/heise medien)
In its news program “heute journal”, ZDF broadcast a report on Sunday evening with AI-generated video material without labeling it. The broadcaster initially explained this violation of its own guidelines as a “technical error” and has removed the AI material from the report. The editorial team apologizes “in all forms.”
The report deals with a “climate of fear” that the sometimes brutal actions of the US border police ICE against illegal migrants create in society. In the introduction, presenter Dunya Halali points out that numerous videos of ICE operations are circulating on the internet: “Not all of them are real, but very many are.”
ICE operations with AI images
The report by chief correspondent Nicola Albrecht then discusses “brutal ICE operations” illustrated with various recordings from the web. “They take children away from their parents,” it says in the originally broadcast report about a video sequence that clearly shows the watermark of OpenAI's generative AI Sora.
Apart from the watermark, the scene, in which alleged ICE employees appear to be separating a mother from her children, contains numerous indications of the use of AI. In addition to the AI-typical blurring of details, these include a fabricated official crest, a generic uniform, and the fact that the alleged ICE employee shows his face.
ZDF speaks of a “technical error”: In the report, “AI-generated images should have been labeled,” according to a correction published by the broadcaster on Monday. “This labeling was not transmitted during the transfer of the report due to technical reasons.” Why ZDF showed AI-generated videos in its report is not explained in the statement.
The broadcast with the corrected report is now available again in the media library. Instead of adding the missing labeling, “heute journal” has removed the questionable scene and partially replaced it with other images. Another sequence, showing a minor being taken away by local police forces in the US state of Florida and recorded in 2022, has also been removed by ZDF. A caption now indicates at this point that the “video was subsequently changed for editorial reasons.”
Videos by heise
“Technical Error”
ZDF initially sticks to its explanation that it was a technical error. “The report shows that a climate of fear is created with both real and AI-generated images. We regret that it was not made clear enough which images are real and which are not,” a spokesperson told heise online. However, the report itself does not address potentially fake images.
On Tuesday afternoon, the editorial team also admitted the error and apologized. “This sequence should not have been used without labeling and without context according to ZDF's rules. Another sequence, in which a child is taken away, is real, but comes from a different context from 2022,” the broadcaster said in a statement.
“We apologize in all forms for these errors,” explained deputy editor-in-chief and news director Anne Gelinek. “The report does not meet our standards and should not have been broadcast in this form. We have therefore removed it from all platforms and will report correctively on ZDFheute and in the same place in 'heute journal' this evening.” The broadcaster has not yet answered questions about further consequences.
(vbr)
