Sweden’s Government Offices has blocked employee access to ChatGPT and Claude on official work computers, directing civil servants to use Microsoft Copilot as the sole approved artificial intelligence tool. The restriction, which took effect Monday, is aimed at closing information security gaps exposed by the unregulated use of consumer AI services.
The block applies to the browser versions of OpenAI’s ChatGPT and Anthropic’s Claude on staff service computers, according to reports from Sveriges Radio Ekot and confirmed by broadcaster SVT.
The Government Offices, which supports the Prime Minister and the cabinet ministries, has not issued a total ban on the underlying technology; some employees may still reach the two services through an internally developed AI platform.
At the center of the decision is a formal warning from the Government Offices’ director of digitalisation, Mikael Wettercrantz. In a post on the agency’s intranet, Wettercrantz stated that using AI services without a formal agreement could have “consequences for our information security”.
The concern is not theoretical. An internal audit at the Swedish Transport Administration earlier this year identified roughly ten AI services in widespread use and documented what auditors described as “problematic use.” The review cited both “the sensitivity of the information that has been fed in” and “how the AI-generated responses have been handled”.
Consumer AI tools typically transmit user prompts to servers operated by their providers. Without a formal contract, government agencies have limited control over how that data is stored, processed, or potentially accessed by foreign authorities under legal requests.
Microsoft Copilot was selected as the approved alternative after the Government Offices conducted a security review and tightened the tool’s settings. The agency declined to disclose the specific changes, citing security considerations.
Anthropic Reveals Claude AI Accessed Georgia Voting Data Without Permission
Who is Alice Huang? The Woman at the Centre of Nvidia CEO Jensen Huang Intelligence Allegations
Six Dead, 15 Injured After Gunman Opens Fire in Poland’s Krakow
Paraguay President Quits as Ruling Party Withdraws Support Amid Banking Scandal
A notable wrinkle in the arrangement: Microsoft added Anthropic’s Claude models to Copilot in early 2026, though those models are disabled by default in the European Union unless administrators enable them. The Government Offices has not confirmed whether Claude models are active within its Copilot deployment, leaving some ambiguity about the practical distinction between the blocked and approved services.
Sweden has previously restricted access to China’s DeepSeek AI service, indicating a broader pattern of scrutiny toward unvetted AI tools.
The Government Offices is not acting alone. Sweden’s Tax Agency and the National Board of Institutional Care have already introduced similar restrictions on consumer AI tools, and the Enforcement Authority is preparing its own decision.
The European Parliament took a more sweeping approach in February, disabling AI features across all work devices – including ChatGPT, Claude, and Copilot – after determining it could not guarantee the security of data processed by those services.
The Swedish approach is less a rejection of AI than an attempt to route its use through contractual and technical controls. As Wettercrantz’s intranet message made clear, the issue is not the models themselves but the absence of formal agreements governing how they handle government information.
