Faultline Faultline Kommando 161

Germany · nd · · 2h

Monitoring | AI on the Kotti: When the Berliner serves as a test subject

Deutsch (original) · Auto-translated to English

Anyone who is out and about at the “Kotti” in Berlin will soon be able to be observed by an “AI” that continues to learn with personal pictures. Photo: dpa | Britta Pedersen Berlin pays a private company money to train an “artificial intelligence” using recordings of passers-by. This recently became public through a request from the Left in the House of Representatives.

The location of the data theft is the Kottbusser Tor in Kreuzberg. Dozens of video surveillance devices have already been installed there and will be operated with behavioral scanners in the future. The Black-Red pilot project is expected to start in the fourth quarter of this year, according to a response from the Senate Department of the Interior to a request from the Greens. According to the Senate, the system is not yet operational. The behavioral scanners analyze movement patterns in real time and evaluate the people filmed for perceived danger. Around 4.6 million euros are budgeted for this by 2030.

What is unique in Germany is that the state contract goes to a private company, the IT holding Adesso SE. The company works with video AI provider Staige, in which it has a stake. According to the Senate, the video data remains on police servers, which retain “data sovereignty”. After authorization from the police, the company is allowed to test and retrain the software there. There is “basically” no transmission to third parties.

Anyone who moves in public spaces must expect that personal data will be used to train the AI, Niklas Schrader tells »nd« in relation to training the AI ​​with personal data. The domestic policy spokesman for the left-wing faction in the House of Representatives has asked the Senate to what extent Berliners are becoming training people for commercial private AI software companies.

Which personal data may specifically be used remains unclear. The crucial paragraph 42d in the General Security and Order Act (Asog) states that non-anonymized data may also be used if the “purpose of the test or training with anonymized data is not achieved.” How often this will be the case in practice remains to be seen.

According to Schrader, the paragraph is far too vague. “The current regulation undermines the deletion period and the earmarking,” said the MP. Purpose limitation in data protection law states that personal data may only be collected for specific purposes and may not be later processed in a way that is incompatible with these purposes. According to the Senate Department of the Interior, raw videos can be stored for one month. Selected training and test data, however, may be retained for up to two years.

The relevant paragraph may not only apply to video recordings. Schrader also believes it is possible that the training could also affect very sensitive data, such as content from source telecommunications surveillance. This involves secretly recording communications on people's devices using surveillance software, for example on messengers such as Whatsapp.

»Very sensitive data can be affected by AI training.«

Those affected do not need to be informed about training AI with personal data. “Of course you can ask the authorities about this, but that will be very difficult with image data,” says Schrader. Together with the Greens, the Left is taking action against the Asog amendment before the Berlin Constitutional Court. It will probably be several months before there is anything new about this.

The Asog, which regulates, among other things, the rights of the police, was tightened at the end of 2025 despite massive criticism from the opposition, data protection officers and various legal associations. Only then did the behavioral scanners on the “Kotti” and the testing of “AI” with personal data become legally possible.

In an interview with Netzpolitik.org, Berlin data protection officer Meike Kamp criticizes that the Asog lacks “an overall view across all authorities”. Kamp sees the danger that large parts of Berlin's city center can hardly be passed through unmonitored and the possibility of anonymous use of public space disappears.

By the end of the year, the “crime-prone location” (kbO) Warsaw Bridge will also become an AI video-monitored zone. All rooms under permanent video surveillance should be signposted like the weapon and knife ban zones. The behavioral scanners at Alexanderplatz and Görlitzer Park are planned for 2027. The Left, which received the most votes in Berlin in the House of Representatives election, had announced that it wanted to take back police powers for video surveillance and random checks at the kbOs.

The »nd.Genossenschaft« belongs to its readers and authors. They are the ones who make our journalism accessible to everyone through their contribution: We are not backed by a media company, not a large advertising customer or even a billionaire.

With your support we can continue to:

→ report independently and critically → address overlooked topics → give space to marginalized voices → counter misinformation → advance left-wing debates

With “Pay voluntarily” you can take part. You help ensure that this newspaper has a future. So that nd remains.

Read the full story at the source

Source: nd