Not only did the EU Parliament recently decide to do so, the federal government also supports the project: in the future, anyone who receives images or videos on digital devices should be able to tell at first glance whether they were produced using artificial intelligence (AI). The manufacturers of such programs are obliged to label every machine-generated representation. This is what it says in the “AI Act”, the first European regulation for artificial intelligence. However, the dispute about this has not yet been resolved in Berlin. The Federal Minister of Justice is still wrestling with the Federal Minister of the Interior, for example over questions of biometric facial recognition.
After all, the federal government agrees that social media and digital platforms should make it easier to see which representations correspond to reality – and which were created artificially. This should help to distinguish facts from fake news online. “Image material that was produced using artificial intelligence should, in my opinion, be identified as such. Because images convey the feeling of authenticity. This can be misused,” Justice Minister Marco Buschmann (FDP) just told the Germany editorial network.
The Liberals wanted to spare influential AI manufacturers
Buschmann’s stance is not new; Germany negotiated the “AI Act” with the EU member states over many months. On the one hand, it is intended to contain the risks of artificial intelligence and, on the other hand, to promote the future prospects of digital speech recognition and image programs. Notifications from the office, for example, could then increasingly be created and sent by machines. At the same time, civil rights must also be protected and those affected must be given the opportunity to object to decisions made by machines.
Buschmann supports the EU’s plan – but not unreservedly. In Buschmann’s eyes, the regulation for AI products, which was passed by the European Parliament in mid-March and is due to come into force in May, is an opportunity to reduce fears of new technology and state surveillance, according to his company. On the one hand. On the other hand, the FDP has slowed down in Brussels. She wanted to prevent too many reporting and transparency requirements for AI companies and start-ups. The Liberals also wanted to spare manufacturers of influential programs such as Chat-GPT from the AI regulation, but in vain. Companies in the EU should now be divided into risk classes depending on how great their influence on civil and human rights, elections or vital infrastructure can be.
With the planned labeling requirement for images and videos that come from artificial intelligence, manufacturers must configure their programs in such a way that the AI origin of the images is immediately recognizable to users, with a type of watermark. Buschmann emphasized that this does not stand in the way of creative work, but it does make it clear whether a picture is “more of a work of art or a depiction of reality.” Subsequent digital manipulations that private individuals have carried out on images and then sent them are not affected by the new regulation.
Faeser wants more biometric remote identification – Buschmann doesn’t
It is still controversial in the federal government where artificial intelligence can help fight crime – for example when monitoring public spaces. The EU’s AI regulation prohibits “social scoring”, i.e. people being monitored based on biometric data, as in China, and classified into political or social categories depending on their behavior patterns. However, biometric facial recognition in real time, i.e. the identification of people in public places or at train stations, is permitted in the EU – but only in exceptional cases, for example in law enforcement.
National governments must now clarify exactly what these exceptions look like. In Germany, Interior Minister Nancy Faeser (SPD) is reportedly pushing for more access rights for security authorities. To combat terrorism, for example, biometric remote identification should be used more extensively in public spaces. Justice Minister Buschmann, on the other hand, is slowing down, he fears for basic and civil rights. When asked on Wednesday what would happen next with facial recognition, his spokesman indicated that there was still a need to speak: “We will first check that and then we will see how we progress with the legislative process here in Germany,” he said. A spokeswoman for the Federal Ministry of the Interior declined to comment. We’re still coordinating internally.