Germany Moves to Let Agencies Train AI on Asylum Application Data

Germany's federal government has approved a draft law that will, for the first time, give migration authorities an explicit legal basis to use artificial intelligence when processing asylum, visa and residence applications. Critics, including the organization AlgorithmWatch, say bluntly that people seeking protection are being turned into guinea pigs for unproven AI systems.

What the law provides

The draft comes from Germany's Federal Ministry of the Interior. KIMVG is meant to create a legal framework, including on personal data protection, for developing and deploying intelligent analytical systems within migration administration. The government justifies the move by pointing to persistently high application numbers and limited staffing at the relevant agencies.

The law is also meant to standardize administrative practice across Germany, where individual states and local offices have so far handled similar cases differently. To that end, AI systems would analyze patterns in already-completed asylum proceedings in order to standardize evaluation criteria.

Scope of agency powers

Under the draft, the powers would extend beyond the Federal Office for Migration and Refugees (BAMF) to cover local foreigners' offices, the Foreign Ministry's visa-issuing arm, as well as police, security services and employment agencies handling foreign nationals.

The most controversial element is the authorization for automated internet searches, including publicly accessible social media profiles on platforms such as Instagram, LinkedIn or Facebook, in search of material that could confirm or undermine the credibility of an application.

Human rights groups push back

AlgorithmWatch accuses the government of enabling the systematic use of sensitive personal data from migration processes without sufficient limits, differentiation or transparency. Amnesty International and Pro Asyl have raised similar objections, warning of the risk of algorithmic bias and of overburdened staff mechanically deferring to AI recommendations.

The government plans to deploy unreliable and discriminatory AI systems precisely in asylum procedures, where decisions about protection and people's fates are made - Sarah Lincoln, Gesellschaft für Freiheitsrechte

This is where the alarm bells should be ringing - Clara Bünger, Die Linke

Risk of self-reinforcing bias

Sarah Lincoln, a lawyer with the Gesellschaft für Freiheitsrechte, notes that deploying unreliable AI systems specifically in asylum procedures means making decisions about protection and people's fates using tools whose errors have not yet been adequately studied. Clara Bünger, a member of parliament for Die Linke, warns that AI reinforces existing biases, and that groups already vulnerable to discrimination will be scrutinized more intensively, without enough officials able to critically assess the algorithm's recommendations.

Experts point to a self-reinforcing bias mechanism: if an AI system flags as suspicious applications from people of certain countries of origin or universities previously linked to fraud cases, those groups will be checked more often, which will statistically generate more detected irregularities and entrench the pattern, regardless of the actual guilt of individual applicants.

What happens next

The bill has already cleared approval by Friedrich Merz's cabinet and now moves into the parliamentary procedure. Debate in the Bundestag and Bundesrat is scheduled for autumn 2026, after the summer recess, leaving NGOs and the opposition time to keep pressing on the provisions covering data protection and appeal mechanisms.

For Polish readers, the story matters beyond Germany's borders. As a major transit and destination country within the European Union, Germany is testing approaches that could later serve as a model for other member states considering the automation of migration procedures, including Poland, where the use of AI in public administration is still in its early stages.