The Swedish government has moved to restrict access to two of the world's most widely used artificial intelligence platforms, ChatGPT and Claude, over concerns about the protection of sensitive government data.
The measure came into force on 26 October and applies to the web-based versions of both services on computers used by employees in government offices and ministries.
According to reports by Euronews, the decision does not amount to a blanket ban on artificial intelligence in Sweden. Instead, it limits the use of specific AI applications within central government administration.
What prompted the decision?
The restrictions followed an audit carried out earlier this year at the Swedish Transport Agency, examining how employees were using AI tools in their daily work.
The review identified cases of problematic use, raising concerns about the handling of information not intended for public disclosure.
Swedish authorities' main concern is the possibility that employees could enter sensitive official data into external platforms without sufficient contractual safeguards governing how that information is stored and processed.
In such circumstances, data could potentially be processed or retained in ways that state authorities cannot adequately supervise.
Mikael Wetterkrantz, director of digitalisation for Sweden's Government Offices, warned that the use of unapproved applications could create serious information security risks.
The issue is commonly referred to as "shadow IT" - the use of digital services by employees without approval or oversight from official IT departments.
Microsoft Copilot becomes the approved alternative
Despite the restrictions, Sweden is not seeking to prevent public employees from using artificial intelligence.
Instead, authorities are promoting the use of Microsoft Copilot, which has undergone security reviews and is covered by an agreement with the Swedish state.
The aim is to preserve the productivity benefits of AI while reducing the risks associated with handling confidential information.
ChatGPT and Claude will not disappear entirely from the government system. A limited number of employees will still be permitted to use the tools through a dedicated and controlled AI platform.
Similar restrictions elsewhere
The move is not an isolated one.
According to Sveriges Radio, similar restrictions have already been introduced by Sweden's Tax Agency and the National Board of Institutional Care.
Meanwhile, the Swedish Enforcement Authority is reportedly considering whether to adopt a similar policy.
The list of non-approved AI applications already includes the Chinese platform DeepSeek, which has also attracted scrutiny over data security concerns.
