According to Sveriges Radio, Regeringskansliet, the administrative body supporting the Swedish government, has decided to prohibit the use of OpenAI's ChatGPT and Anthropic's Claude on its work systems. The move reflects growing concern over how sensitive government information is handled, at a time when several European public administrations are reassessing the risks posed by generative AI tools hosted abroad.
The exact scope of the decision, its implementation timeline, and the specific technical justifications behind it have not been fully detailed in the available reporting. It appears, however, that the central concern relates to how data is processed and stored by US-based providers, a recurring issue for public institutions focused on digital sovereignty and compliance with European data protection rules.
Sweden's move is not an isolated case. Several countries and public bodies have already restricted or regulated the use of consumer-grade generative AI tools within their services, including Italy, Germany, and various private companies concerned about protecting trade secrets. These restrictions reflect an ongoing tension between the productivity benefits promised by conversational assistants and the risks associated with confidentiality, particularly when the servers processing requests are located outside the European Union.
The Swedish case illustrates a broader dynamic: as generative AI tools become embedded in everyday professional workflows, public administrations are forced to weigh rapid adoption against regulatory caution. Absent a clearly identified sovereign alternative, such bans are likely to multiply unless providers offer stronger guarantees regarding data localization and processing for government use cases.