fact finder
Fake, AI-generated audio files from the tagesschau are currently being distributed – including at a demo. The false audios give the impression that the Tagesschau is apologizing for alleged lies in its reporting.
“Good evening, ladies and gentlemen. I welcome you to the Tagesschau. Today we would like to apologize to you. We have been brazenly lying to your face for over three years,” says one of the fake ones daily News-Audio files. The audio starts with the official one daily News-Jingle that can be heard before every broadcast, suggesting that these are recordings of daily News-speaker Susanne Daubner as well as daily News-Spokesman Jens Riewa acts.
A total of three audios are circulating, all of which contain supposed apologies about alleged “deliberate manipulation” and “lies” in the reporting. Thematically it’s about the war in Ukraine, the corona pandemic and the “denunciation” of demonstrators. That’s how I have it daily News lied when they “denounced decent citizens from the middle of society as right-wing extremists, Reich citizens or Corona deniers”. The daily News has been lying to people since the beginning of the war in Ukraine “in order to be able to blame Putin alone for the escalation” or has spread lies about the Covid-19 vaccinations against his better judgment.
In all three audio files it says towards the end: “For all this one-sided reporting and conscious manipulation, especially for the denunciation of our fellow human beings, we must expressly apologize on behalf of public broadcasting.”
Clear fake
Marcus Bornheim, editor-in-chief of ARD-aktuell, makes it clear that these files are fakes. The daily News As the best-known news program in Germany, it is considered serious, credible and reliable. “This is being exploited here to specifically spread disinformation,” says Bornheim. It is absurd that people who describe the press as the “lying press” consciously influence their audience with such fake audio.
If you listen carefully, you can easily reveal these AI-generated audio files as fakes. The voices of the daily News-Speakers sound mechanical, some intonations of the words are incorrect and the speech melody is different from the real speakers. This can be heard particularly well in the transitions at the end, where the speakers’ actual farewells were inserted.
AI-generated audio is often difficult to recognize
However, it is often no longer as easy as these examples to distinguish real from fake audio files. Overall, it is already difficult to recognize AI-generated audio, says Andreas Dengel, Managing Director of the German Research Center for Artificial Intelligence (DFKI): “The AI can also simulate slight nuances of humanity, such as emotions or subjectivity. And then “It makes it even harder to identify something like that.”
From Dengel’s point of view, people must, above all, check the plausibility of the content. “It is very unrealistic for a news channel to publish such audios. Common sense should intervene and say that these audios cannot be real.” However, it also depends very much on who hears such a fake. “The better a message fits into people’s existing worldview, the more likely it is to be believed.”
Audios played on demonstration
It is difficult to say whether the audio files will be viewed as credible by the people who hear them or not. However, according to a live video stream, the three audios were probably played via loudspeaker announcements at a demonstration in Kiel at the beginning of November, which was registered under the name “The Great Mutiny”.
According to the “Kieler Nachrichten”, the “Kiel Yellow Vests”, which emerged from the protests against the Corona policy, mobilized for the demonstration across Germany – around 1,500 people came. The participants are primarily people from the so-called lateral thinker scene, says Christian Martin, Professor of Comparative Politics at the Christian Albrechts University in Kiel. “Politically, they cannot be clearly assigned to one direction; sometimes things are wildly mixed up. As a bracket, you can say that people feel neglected and believe that the state is restricting their freedoms.”
demonstrators from the Lateral thinking milieu
In terms of content and people, the demo can be compared to the protests against the Corona measures during the pandemic, says Martin. “There is the idea of an evil elite that is responsible for all the evil in the world.” For example, one of the speakers was the Hamburg doctor Walter Weber, against whom the Hamburg public prosecutor’s office brought charges last year because he is said to have issued false certificates for mask exemption.
On the demo’s website you can find positions similar to those in the AI-generated audios. In addition to the demand for an allegedly necessary separation of the state and the media, which would allow themselves to be bribed by politicians for their reports, the demands range from a stop to arms deliveries to Ukraine to conspiracy stories surrounding the World Health Organization and the United Nations.
The site’s imprint lists Henry Gienap, who ran for the AfD in this year’s municipal elections in Kiel. From Martin’s point of view, this is not surprising: “People who feel comfortable in such an environment also have an affinity for the AfD.”
Spread on social networks
The AI-generated audios were also distributed in some relevant conspiracy ideological Telegram channels, on Facebook and TikTok and were shown to several thousand users each – although usually with the indication that it was satire. However, Josef Holnburger from the Center for Monitoring, Analysis and Strategy (CeMAS) points out that labels of satire do not have to be an indication for the audience. “Satirical references are often used specifically as a strategy by disinformation spreaders in order to protect themselves against possible impending charges within the framework of artistic freedom.” This is sometimes more, sometimes less promising.
The content would remain with users, explains Holnburger. “In the conspiracy ideology scene, people are convinced that the media would lie about everything – that’s why such an apology is a wishful thinking for many.” In the end, a large part of the scene doesn’t care whether audio recordings or quotes were actually made that way; it is usually enough to believe that statements could have been made that way.
“Strategies of the delegitimization”
According to Martin, audio files like these need to be taken seriously. “The very fact that such audios exist naturally undermines the credibility of statements made on public broadcasting,” says political scientist Martin. Ultimately, these are all “strategies of delegitimization,” says Martin. An attempt is being made to create confusion so that less attention is paid to the real facts.
Editor-in-chief Bornheim also emphasizes: “The positions and strategies represented in the AI-generated audios are capable of damaging the free-democratic basic order.”
Abuse of the credibility from newscasters
Since it will become increasingly easier to produce AI-generated fakes in the future, Holnburger assumes that these will increase. False content from news anchors is already circulating again and again. For example, AI-generated videos made by tagesschau24Spokesman André Schünke or “ZDF aktuell-journal” spokesman Christian Sievers made the rounds in which they supposedly advertised dubious financial investments. In both cases they are fakes.
“The creators of these fake investment recommendations, also known as scams, often use well-known personalities to give their alleged investment recommendations more credibility,” says Holnburger. The hope of the fraudsters is that the people will get the message: If even they daily News reported about it, it can only be serious.
“The speakers and their programs already have a certain level of credibility among the population. This trust can then be exploited for such purposes,” says Dengel. In addition, there is a lot of material from news anchors with which the AI can be trained. The more data there is, the better the end result will be.