Cracking AI secrecy

In core government and for public tasks hugely sensitive for citizens such as migration governance in Europe, the use of AI systems developed and provided under contract gives migration authorities and private providers the power to keep their deployment secret and to refuse access to their key components. This article explores the complex interplay between public and private drivers of secrecy surrounding procured AI technologies in European migration governance, demonstrating how these forces converge and reinforce one another. It examines AI secrecy as an institutional framework shaped by both public and private law forms of secrecy, applying the spectrum of secrecy from deep (unknown unknowns) to shallow secrecy (known unknowns). Specifically, the paper looks into the sources of ‘political secrecy’ in migration and security authorities, highlighting how migration agencies and authorities keep the existence and use of AI systems from civil society and affected persons (such as third country nationals). Next, it analyses the legal frameworks that sustain AI vendors’ private secrecy, such as trade secrecy, secrecy in procurement procedures and agreements. In examining the various secrecy points, it scrutinises the limited impact of the AI Act in addressing the entrenched secrecy of AI deployment and development in governing migration. Finally, we condense the main takeaways and examine the broader repercussions of AI secrecy beyond the context of border control, touching upon its implications for a new equilibrium transcending the conventional public/private divide.

Civil society · Corporate governance · National security · Secrecy · Software deployment · Cybersecurity and Cyber Warfare Studies · Digital Economy and Work Transformation · Ethics and Social Impacts of AI