Hardware Ready Ready to Bench?
The Swedish government has decided to block access to ChatGPT and Claude on its employees’ work computers, directing them toward authorized artificial intelligence tools, including Microsoft Copilot. The decision is motivated by information security needs and is part of a broader effort to regulate the use of artificial intelligence.
The restriction will take effect on Monday, October 26, and will apply to employees of the Regeringskansliet, the administrative apparatus that supports the Swedish government. The news was reported by Sveriges Radio.
ChatGPT and Claude: why it matters
Mikael Wettercrantz, director for digitalization at the Regeringskansliet, explained in a message published on the administration’s intranet that using AI services without an agreement with the entity can have “consequences for our information security.” The main concern is the potential inclusion of internal documents, sensitive data, or other information requiring specific confidentiality guarantees in prompts.
Without clear rules and appropriately configured tools, the organization risks losing control over how this information is handled. The restriction on ChatGPT and Claude comes with a review of the security measures applied to Microsoft Copilot, which remains among the authorized tools.
According to internal information, the Regeringskansliet decided to strengthen service protection following an analysis. However, the administration did not provide details on the changes introduced, citing the sensitive nature of the information regarding security as the reason for this choice.
The block will not be absolute for all staff. Some employees will continue to have access to ChatGPT and Claude through a dedicated platform for artificial intelligence assistants developed directly by the Regeringskansliet. This approach suggests a distinction between access to public services in their standard configuration and that mediated by an environment controlled by the administration.
What changes and what are the effects
The Swedish government had already introduced restrictions for DeepSeek, a Chinese-origin generative AI service. An investigation conducted by Sveriges Radio reveals that similar restrictions have been introduced at the tax agency Skatteverket and the Statens institutionsstyrelse (SiS), the body managing residential facilities for people subject to specific assistance or judicial measures.
Kronofogden, the Swedish authority responsible for forced execution and debt recovery, is also evaluating the issue ahead of a decision. Another significant case involves Trafikverket, the body responsible for the country’s road and rail infrastructure. An internal audit had detected extensive and risky use of about ten artificial intelligence services, highlighting problems both in the data entered by users and in the management of generated responses.
Source and further reading on ChatGPT and Claude: original article.
* Content created with the assistance of artificial intelligence systems.