Sweden’s Government Offices will block ChatGPT and Claude on staff computers from Monday, steering civil servants to Microsoft Copilot as the only AI tool cleared by its internal security review.
The move signals growing governmental scrutiny of third-party AI platforms, as public administrations weigh productivity gains against data protection and national security risks. Microsoft Copilot emerged as the sole approved option, while OpenAI’s ChatGPT and Anthropic’s Claude were excluded from official use.
Key Facts
- Sweden’s Government Offices will block ChatGPT and Claude on staff computers starting Monday.
- Civil servants will be directed to use Microsoft Copilot as the only AI platform approved after a security review.
- The decision follows an internal security assessment conducted by Sweden’s Government Offices.
- The affected tools include ChatGPT (OpenAI) and Claude (Anthropic).
- The restriction applies solely to staff computers within the Government Offices.
What Happens Next?
The immediate next step is the technical enforcement of the policy, which takes effect on Monday. Staff computers will reportedly have access to ChatGPT and Claude restricted, while Microsoft Copilot becomes the default AI assistant available to civil servants.
The rationale, according to Decrypt, stems from a security review conducted internally by Sweden’s Government Offices. While the summary does not detail the specific criteria used to evaluate each platform, the outcome suggests that Microsoft Copilot met standards that ChatGPT and Claude did not.
This decision aligns with broader European concerns about data sovereignty and AI regulation. As governments across the EU grapple with the rise of generative AI, national security and privacy considerations are increasingly shaping public-sector procurement and usage policies. Whether other European countries will follow Sweden’s lead remains to be seen, especially given Microsoft’s integration with existing government IT infrastructure.
Who Is Affected?
The policy directly affects civil servants working within Sweden’s Government Offices, also known as Statsverket. These employees rely on digital tools for communication, documentation, and administrative tasks, and the shift mandates their exclusive use of Microsoft Copilot for AI-assisted workflows.
Importantly, the restriction applies only to official staff computers, meaning personal devices or non-government networks are not impacted by this directive. However, since many public-sector roles involve handling sensitive data, the policy reflects an effort to standardize and control how AI tools interact with confidential information.
Government employees who previously used ChatGPT or Claude for drafting memos, summarizing documents, or automating routine tasks must now transition to a single approved alternative. This could affect productivity in the short term, depending on how well Copilot integrates with existing workflows and software suites already used by the agency.
What We Know — and What We Don’t
Verified by the source:
- Sweden’s Government Offices plans to block ChatGPT and Claude on staff computers.
- Microsoft Copilot is the only AI platform approved following a security review.
- The change takes effect on Monday.
- The policy targets civil servants using official government devices.
Still unconfirmed:
- The specific security concerns or evaluation criteria used in the review process.
- Whether other branches of the Swedish government will adopt similar restrictions.
- Detailed responses or comments from OpenAI, Anthropic, or Microsoft regarding the decision.
- Long-term impacts on civil servant productivity or policy adoption timelines.
Why It Matters
This development highlights the increasing tension between innovation and oversight in public governance. As generative AI becomes embedded in daily operations, governments face mounting pressure to balance utility with risk. Sweden’s targeted ban reflects a growing trend among democratic states to assert tighter controls over how sensitive data is processed and stored, especially when powered by foreign-developed technologies.
What To Watch
It remains to be seen whether other European administrations will mirror Sweden’s approach or adopt more permissive AI usage guidelines. Additional clarity may emerge if further details about the security review criteria are disclosed.