Berlin will begin AI-assisted video surveillance at Kottbusser Tor in November, Interior Senator Iris Spranger said on Wednesday. Thirty cameras will monitor the square around the clock and automatically flag possible violent crimes or emergencies, with regular operation planned to start in early December after a trial phase.
Kottbusser Tor is one of seven locations Berlin classifies as crime hotspots. Spranger cited more than 900 offences registered in the area over the past 12 months, including more than 300 physical assaults. The scheme is based on a reform of Berlin's General Security and Ordnance Act passed by the governing black-red coalition.
The system analyses video continuously and reduces people to simplified stick figures to recognise punches, kicks or a person lying on the ground. When the AI raises an alarm, officers at a police station decide whether to intervene and see the normal, identifiable video image. Spranger acknowledged during a question-and-answer session that her earlier statement that the images would remain anonymised was not correct.
Recordings are stored for 28 days and can be secured and used as evidence if they show an offence. The data is held in a Berlin police data centre and, according to Police President Barbara Slowik Meisel, is encrypted and secured to BSI standards.
How accurate the system will be in everyday use remains open. In Hamburg, 99 percent of situations flagged by an AI system turned out to be false alarms. Police director Markus van Stegen said Berlin had deliberately chosen a different system and expected significantly better results.
The project's political future is also uncertain. A possible new coalition of the SPD, Greens and Left has agreed in its exploratory paper to evaluate video surveillance, and the Berlin Left has been critical of cameras in public space in the past. Spranger said she would hold to the project, adding that the Greens and the Left also had to guarantee security in Berlin. Further AI cameras are planned for Warschauer Brücke, Alexanderplatz and Görlitzer Park.
Analysis
Why This Matters
- Berlin is putting AI-assisted video surveillance into regular use at a busy public square, testing whether the technology can distinguish real emergencies from ordinary movement.
- The system raises privacy questions: the AI itself works on anonymised stick figures, but officers see identifiable images, and recordings are kept for 28 days.
- The project's fate is tied to ongoing coalition talks, so a change of government could change Berlin's surveillance policy.
Background
Kottbusser Tor is one of seven locations Berlin officially classifies as crime hotspots. The legal basis for the AI surveillance was created by a reform of Berlin's General Security and Ordnance Act passed by the state's black-red coalition. Hamburg's experience, where 99 percent of AI-triggered situations were false alarms, is a key reference point for doubts about the technology's reliability.
Key Perspectives
- Iris Spranger (SPD, Berlin interior senator): says the system is needed because more than 900 offences were registered around Kottbusser Tor in the past year, including more than 300 physical assaults, and it should help recognise dangers earlier and protect people.
- Markus van Stegen (Berlin police director): expects significantly better results than in Hamburg because Berlin deliberately chose a different system.
- Berlin's Left (Die Linke): has been critical of surveillance cameras in public space in the past, and the SPD, Greens and Left have agreed in an exploratory paper to evaluate video surveillance.
What to Watch
- Whether coalition negotiations between the SPD, Greens and Left change, delay or dismantle the AI surveillance project.
- The false-alarm rate once the system moves into regular operation in early December.
- Whether the planned expansion to Warschauer Brücke, Alexanderplatz and Görlitzer Park goes ahead.
Checked
11 claims in this story checked against 1 source: 10 supported.
Show each claim
- SupportedBerlin will begin AI-assisted video surveillance at Kottbusser Tor in November.Politico Europe: “Im November startet am Kottbusser Tor in Berlin die KI-gestützte Videoüberwachung.”
- SupportedThirty cameras will monitor the square around the clock.Politico Europe: “30 Kameras sollen den Platz rund um die Uhr erfassen und mögliche Gewalttaten oder Notfälle automatisch erkennen.”
- SupportedRegular operation is planned to start in early December after a trial phase.Politico Europe: “Nach einem „Probe-Echtbetrieb“ soll das System Anfang Dezember in den regulären Betrieb gehen.”
- Partly supportedSpranger said on Wednesday that Berlin’s system would begin in November.Politico Europe: “Innensenatorin Iris Spranger (SPD) begründete den Schritt bei der Vorstellung am Mittwoch mit der Kriminalitätslage”
- SupportedKottbusser Tor is one of seven locations Berlin classifies as crime hotspots.Politico Europe: “Das Kottbusser Tor ist einer von sieben sogenannten kriminalitätsbelasteten Orten in Berlin.”
- SupportedMore than 900 offences were registered in the area over the past 12 months, according to Spranger.Politico Europe: “In den vergangenen zwölf Monaten seien im Bereich Kottbusser Tor mehr als 900 Straftaten registriert worden, darunter mehr als 300 körperliche Angriffe.”
- SupportedMore than 300 physical assaults were registered there over the past 12 months, according to Spranger.Politico Europe: “In den vergangenen zwölf Monaten seien im Bereich Kottbusser Tor mehr als 900 Straftaten registriert worden, darunter mehr als 300 körperliche Angriffe.”
- SupportedThe system analyses video continuously and uses simplified stick figures to identify punches, kicks or someone lying on the ground.Politico Europe: “Die KI analysiert das Videomaterial fortlaufend auf bestimmte Bewegungsmuster. Dafür abstrahiert sie die abgebildeten Personen vereinfacht zu einer Art „Strichmännchen“ und wertet deren Bewegungen aus. So soll sie etwa Schläge, Tritte oder eine am Boden liegende Person erkennen.”
- SupportedAfter an alarm, police at the station see identifiable normal video images and an officer at the monitor decides whether to intervene.Politico Europe: “Ob die Polizei tatsächlich eingreift, entscheidet der Beamte am Monitor. ... Die Beamten sehen nach einem Alarm das normale Videobild, auf dem Personen zu erkennen sind.”
- SupportedSpranger acknowledged during the question-and-answer session that her earlier statement that images would remain anonymised was incorrect.Politico Europe: “Auch Spranger räumt in der Fragerunde ein, dass ihre vorherige Darstellung – die Bilder blieben anonymisiert – so nicht richtig gewesen sei.”
- SupportedIn Hamburg, 99 percent of situations flagged by an AI system turned out to be false alarms.Politico Europe: “In Hamburg haben sich 99 Prozent der von der KI ausgelösten Situationen als Fehlalarme erwiesen.”
