Digital Technology
and Global Security

Where do AI platforms emerge from? And which practices, people, and places enable them? The project Sites of AI Empire is an interactive map that critically examines the power structures and human resources enabling AI systems used in public administration, such as the large language model LLMoin used in Hamburg. While LLMoin is presented by the city authorities as an “efficient” and “secure” AI tool for administrative tasks, this project instead explores the often invisible and overlooked social, material, and immaterial conditions behind these systems. With the help of the online mapping tool Kumu, it situates LLMoin within an exemplary global network of the AI industry, while drawing on theoretical and critical perspectives from Science and Technology Studies to AI Empire.

How does AI imagine its own relationship to the natural world — and what does that reveal about the systems we are building? This project developed an innovative methodology for "interviewing AI": systematically prompting ChatGPT, Grok, and DeepSeek to generate images on the topic of AI and nature, then analysing the results for recurring motifs, framings, and ideological assumptions. The findings are striking — all three systems consistently produced idealised, harmonious visions of technology and nature coexisting, despite the well-documented ecological costs of AI infrastructure in terms of energy, water, and raw material consumption. The comparative analysis also revealed distinct differences between the systems, pointing to biases embedded in training data and model architecture. The project argues that AI-generated imagery is never neutral: it encodes and reproduces particular assumptions about progress, technology, and the environment.

What makes a public space feel secure — or insecure? And who, or what, produces that feeling? Drawing on Actor Network Theory and the "making and doing" approach in Critical Security Studies, this project traced how security is assembled through the interactions of cameras, lighting, barriers, signs, sanitation infrastructure, security personnel, and the people who move through these spaces. Through repeated field visits, non-participant observation, and visual ethnography at Hachmannplatz and Hansaplatz in Hamburg, the group mapped these actor networks digitally — making visible what usually remains invisible. The result is an interactive online map that documents how security is not a stable condition but a contingent, ongoing achievement — one that varies by time of day, shifts with social context, and frequently operates through subtle exclusion.

Drawing on Critical Security Studies, this project examined AI-based video surveillance as a form of urban governance — not merely a technical tool for crime prevention. Through document analysis and a mini-survey conducted at Hansaplatz and Hachmannplatz in Hamburg, the group found that AI surveillance shifts security logic from reactive policing to the anticipatory detection of "suspicious" behaviour, with automated systems effectively co-defining what counts as normal or deviant conduct in public space. Public perceptions proved ambivalent: while a majority perceived greater safety, significant concerns about intrusive monitoring and loss of control were also evident. The project concludes that AI surveillance represents a socio-political transformation of urban governance — one that reshapes power relations and raises fundamental questions about civil liberties and democratic accountability.

What happens when the decision to grant or deny asylum is delegated to an algorithm? This podcast episode follows an imagined migrant arriving at the EU's external border — and uses that narrative to unpack the real and growing role of AI in European border and migration control. The episode examines how borders have become zones of exception, where asymmetric power relations enable unprecedented levels of technological experimentation. At the centre of the analysis is the EU-funded iBorderCtrl project, which deployed AI-based lie detection and emotion recognition at border crossings — techniques that have faced withering scientific critique for their unreliability and bias. Drawing on recent scholarship (Monedero & Dencik 2022; Rinaldi & Teo 2025; Vavoula 2024), the episode traces how these experimental systems translate into concrete human rights harms — risk of discrimination, denial of asylum, and lack of transparency — and reflects on the shortcomings of the EU's emerging AI governance framework in addressing them.