Frontex’s proposed return chatbot shows how an information service can pursue a policy outcome

The planned app would give people information about returning to their countries of origin, including financial support and further counseling. Frontex also hoped it would increase the number choosing return.

That institutional objective complicates the description of the tool as neutral information delivery. The reporting, drawing on documents obtained by AlgorithmWatch, also identified uncertainty about training material and whether intended languages such as Arabic, Urdu, and Pashto could be adequately supported.

Frontex’s internal review did not classify the system as high risk under the AI Act. The article questions what that classification leaves unexamined, especially when advice may influence consequential decisions by people with limited alternatives. This is a February account of a project under development, not evidence that the chatbot subsequently delivered accurate multilingual counseling.