The deputy head of the Federal Press Office (BPA), Johannes Dimroth, considers the state resources with which democracies can react to deepfakes to be inadequate. Relevant image, audio or video manipulations that are generated with the help of artificial intelligence (AI) undermine the “trust in the authenticity of content as a whole,” he explained on Friday at a round table on “Trustworthy AI” organized by the Tagesspiegel in Berlin. An individual case can, if necessary, be solved using a “red telephone”. In the long term, however, an anchor for democratic opinion formation is in danger. Media literacy and “pre-bunking”, i.e. immunization against disinformation by putting on mental armor, are important building blocks in the fight against deepfakes. The lawyer also doesn't want to let the tech companies off the hook.
Advertisement
Dimroth stated that the federal government has not yet experienced the “big bang” that could undermine the integrity of elections. However, there are attempts every day to influence the opinion-forming processes of local politics and the public. Chancellor Olaf Scholz (SPD) himself is considered a burned child: In November, the “Center for Political Beauty” (ZPS) had an AI-animated head of government announce in a video speech that it would soon apply for a ban on the AfD before the Federal Constitutional Court. Dimroth explained that the artist collective's “remarkable hard work” was visible on a large screen as AI-generated content. “But not on the subway on a cell phone.”
For the speaker, one thing is certain: Wherever the federal government is written on it, it must be ensured that it is there. It is not acceptable for the Chancellor to apparently tell the interested public things that he does not want to do. “I'm talking tough about security policy interests in multiple crises,” emphasized Dimroth. In the ZPS case, the BPA first used the “notice and take-down procedure” on platforms such as Instagram or YouTube and requested that the video be taken down. However, the operators traditionally find it very difficult to do this because they are forced into an arbitrator role and have to make decisions based on aspects such as voluntary commitment, copyright or trademark law and the threat of confusion.
Anti-vaxxers and the “perversion” of freedom of expression
The considerations and the further procedure “took a very long time,” complained the former head of the BPA’s political communication department. Some platforms took down the content, others not. At the same time, the ZPS, as the author, threatened legal proceedings with the aim of condemning the government to refrain from taking action. They then sought a decision in court themselves. Only after seven weeks in the interim legal protection proceedings did the judges recognize a risk of confusion that could not be justified by art or satire. Nevertheless, the video can still be found in various places on the Internet.
“The instruments are actually not adequate,” Dimroth sums up after the “steep learning curve.” This is an organization that stands by the deepfake and promotes it. “That actually increased our concern” with regard to other, less open initiators, the expert reported. Ultimately, everyone on the Internet is a broadcaster and many such loudspeakers can “claim trust among certain population groups.” At the same time, tech-driven by algorithms, many listeners felt their opinions were confirmed. It is becoming increasingly difficult to “bring people back”. Fewer and fewer people could tolerate a counter-speech, and more and more said goodbye to democratic discourse.
According to Dimroth, the corona vaccination debate is an example of this. Opponents of immunizations have complained loudly that they are no longer allowed to say what they think. However, Article 5 of the Basic Law in no way protects citizens from opposition and is not restricted by the state. Rather, anti-vaccination opponents were engaging in a “perversion of the basic idea of Article 5.” Collective trauma also played a major role. The current solidarity of the Global South with Palestine, for example, “we cannot successfully address with our point of view.” This is where different worldviews come together after the North has also destroyed trust.
The new order of truth and its limits
“Can we even perceive truth anymore?” asked Jürgen Neyer from the Social Science Lab for Artificial Intelligence and Research-based Learning at the European New School of Digital Studies in Frankfurt (Oder). “We are socialized to believe in the strength of good arguments.” This held society together for a long time. But it is becoming increasingly clear that narrative integration is becoming more important. The social scientist explained that people lived “in certain orders of truth”: “Everything breaks down into little stories. We can no longer clearly distinguish between lies and truth. This leads to great uncertainty.” He therefore advocates a “probabilistic understanding of truth”. This is also how AI works, which by no means makes objective decisions. A dichotomous black and white picture of the world is outdated.
At the same time, Neyer advocated not making the issue of regulating deepfakes too strict, for example via the Digital Services Act (DSA). He warned against overblocking, especially content that would be perceived as “disruptive to public order in the future.” Such blockages had great potential for abuse. In addition: “Critical discourse makes us resilient.” It is also important to deal with “the ugly, the unpleasant.” Technical aids such as signatures or watermarks are of little help against deepfakes: they can be forged.
Wolfgang Dierker, head of the legal department at Microsoft Germany, nevertheless praised the agreement reached at the Munich Security Conference by 25 companies such as Adobe, Anthropic, OpenAI, Meta, Google, Amazon, Snap, of AI in democratic elections. However, he admitted that this could only represent part of the answer and that “collaboration between all social players” including training and awareness-raising was necessary in order to “get ahead of the wave”. At the same time, the lawyer made it clear: “You have to be able to trust in authenticity. This is where the debate about a new order of truth ends.”
(bme)