AI in migration management: 'It's about making decisions about people's lives'

Germany’s federal government has reached agreement on a proposed law aimed at integrating artificial intelligence into its migration administration, a move that critics say could have profound implications for human rights and data privacy. The legislation, finalized in late July, would expand the use of AI in asylum and residence permit procedures, affecting not only asylum seekers but also individuals applying for visas or holding temporary permits such as tolerances or residency authorizations. According to the draft, authorities handling these matters would be permitted to use personal data to train and test AI systems, implement automated monitoring of processes, and cross-check information from applicants with online sources. The law introduces three key changes. First, it allows immigration-related agencies to use personal data to develop and test AI tools. Second, it enables the creation of “automated process monitoring” mechanisms, systems designed to detect patterns indicating whether certain groups of applicants are disproportionately rejected based on similar characteristics. Third, it grants the Federal Office for Migration and Refugees (BAMF) and local foreign affairs offices the right to automatically verify applicant statements against internet sources when doubts arise. Compared to earlier drafts, the revised version includes minor amendments, such as explicitly prohibiting the use of biometric data in automated checks conducted by BAMF and local authorities. However, this change does not fully address concerns raised by advocacy groups and federal ombudsmen. For example, Ferda Ataman, Germany’s federal anti-discrimination commissioner, expressed clear reservations during discussions over the bill. She criticized the lack of emphasis on banning discriminatory algorithms, noting that the draft merely requires agencies to ensure that “discriminatory algorithms are neither developed nor used.” This is to be achieved through “technical and organizational measures,” which she argued were insufficient. Ataman called for annual transparency reports detailing the AI systems being developed and deployed. Similarly, several organizations, including AlgorithmWatch, have voiced concerns about the adequacy of safeguards against discrimination. Nora Oppermann, a representative of AlgorithmWatch, warned against exposing individuals to immature AI systems and using their life circumstances for research lacking empirical evidence. Pro Asyl, a prominent advocacy group, stated in its position paper that the provisions regarding system traceability, protection of sensitive personal data, and prevention of discriminatory decisions are inadequate. Wiebke Judith, a legal policy spokesperson for the organization, emphasized in a press release that the proposed law represents a “blank check” for the application of AI across numerous government bodies. She stressed that the issue at hand involves “decisions about human lives.” The group urged the German parliament to either abandon the proposal or substantially revise it. Parliamentary debates on the draft are set to resume after the summer recess. Louisa Specht-Riemenschneider, the federal commissioner for data protection and freedom of information (BfDI), also called for revisions to the draft. A spokesperson for the agency told netzpolitik.org that while they acknowledge the potential efficiency gains from AI use in administrative processes, these benefits must not come at the expense of constitutional principles. The BfDI highlighted the need for clearer regulations specifying which data can be used in the development and deployment of AI systems. Where anonymized data suffices, the use of personal information should be avoided.