Sweden’s restrictions on ChatGPT and Claude took effect on Monday, October 26, 2026, preventing government employees from using the browser-based versions of the artificial intelligence tools on their work computers over concerns about how sensitive information may be handled.
The decision follows a review of AI use at the Swedish Transport Administration earlier this year, which identified what officials described as “problematic” uses of the platforms in government work. The findings raised concerns about employees entering sensitive information into AI services without agreements that clearly establish how the data will be protected and processed.
Officials fear that information submitted to AI platforms without adequate safeguards could be retained or used for model training, potentially exposing sensitive government data to risks beyond the control of public authorities.
Mikael Wettercrantz, director of digitalisation at the Government Offices of Sweden, warned that using AI tools outside approved security arrangements could have consequences for the country’s information security. Such unapproved software and services are commonly referred to as “shadow IT”, a term used when employees adopt digital tools without formal approval or oversight from their organisation.
The restrictions do not amount to a complete ban on artificial intelligence in Swedish government work. Employees can continue using AI tools under tighter security arrangements, while a limited number of authorised staff will retain access to ChatGPT and Claude through a special platform for AI assistants developed by Swedish authorities.
The government is instead recommending Microsoft’s Copilot as an alternative. According to officials, the service has undergone a security screening process and is covered by an agreement between Microsoft and the Swedish government that sets out protections for data handling.
The distinction highlights the growing challenge facing public institutions as they introduce AI into everyday administrative work. Although these tools can help employees draft documents, summarise information and complete routine tasks, their use in government settings raises questions about confidentiality, data retention and the conditions under which information is processed.
The concerns are not limited to the Government Offices of Sweden. According to Swedish public broadcaster Sveriges Radio, restrictions on ChatGPT and Claude have also been introduced at the Swedish Tax Agency and the National Board of Institutional Care. The Swedish Enforcement Authority is reviewing the issue.
The two platforms have also joined DeepSeek on the Swedish government’s list of AI tools that are not approved for general official use because of data security concerns. The development reflects a broader effort by public institutions to assess AI services before allowing employees to use them with official information.
The restrictions come as governments and public agencies seek ways to benefit from artificial intelligence without compromising sensitive records. For Swedish civil servants, the immediate change is that access to popular AI assistants will depend on approved systems and security arrangements rather than individual choice.
The policy leaves room for AI use, but places greater emphasis on where government data goes, how it is handled and whether the services employees use meet the security requirements set by their institutions.
