Swedish government offices are cutting off access to ChatGPT and Claude on their staff computers starting Monday, October 26. Only Microsoft Copilot remains authorized, after a security review. Except that Copilot can run Claude models.
In brief
- Swedish government offices ban ChatGPT and Claude to limit data risks.
- Microsoft Copilot remains authorized after reviewing and strengthening its security settings.
- Other Swedish administrations are taking similar measures in response to the growing use of AI.
Data security, why Stockholm is tightening the screw
This decision is part of a broader trend. Several administrations are seeking to regulate the use of artificial intelligence, as illustrated by the withdrawal of DeepSeek from American federal agencies. Italy had also blocked ChatGPT in 2023.
In Sweden, public radio revealed the measure, confirmed internally. Mikael Wettercrantz, director of digitalization for government offices, explained on the intranet that using AI services without a formal contract could compromise information security.
The services noted in particular problematic uses, linked to the sensitivity of the data transmitted and the processing of the responses generated.
The problem therefore concerns as much the conditions of use as the models themselves. With a consumer service, agents transmit their requests to an external provider, without always controlling data processing. A contract, on the other hand, allows guarantees and responsibilities to be defined.
This is precisely what sets Copilot apart from the two banned tools. After reviewing Microsoft’s solution, the administration strengthened its security settings before authorizing it. However, some agents can access ChatGPT and Claude via an internal AI platform. The press service did not specify how many people benefit from it or under what conditions.
Other Swedish administrations follow suit
The restriction does not only apply to government offices. The Fiscal Agency and the Institutional Care Board have already adopted comparable measures. For its part, the Executing Authority, responsible in particular for debt recovery, is preparing its own decision.
Furthermore, an internal audit of the transport administration identified around ten AI services used widely and presenting risks. These findings illustrate the gap between the rapid adoption of these tools and the implementation of appropriate rules.
Other European institutions are also moving in this direction. In February, the European Parliament disabled AI functions on its professional devices, including those of ChatGPT, Claude and Copilot, due to lack of sufficient data protection guarantees. In the United States, Anthropic lost its appeal to the Pentagon in September, amid tensions around the use of its technologies.
Authorized copilot, Claude potentially present
The distinction made by Stockholm, however, deserves closer examination. Microsoft integrated Claude models into Copilot in early 2026. In the European Union, their activation requires voluntary intervention from the administrator, as they are disabled by default.
However, Swedish government offices have not indicated whether their version of Copilot allows these models to be used. The ban on Claude and the authorization of Copilot could therefore coexist without the technical boundary being as clear as it seems.
In reality, Stockholm seems to favor tools whose administration controls the conditions of use, rather than banning a technology as a whole. It remains to be seen what precise guarantees justified Microsoft’s choice.
Maximize your Tremplin.io experience with our ‘Read to Earn’ program! For every article you read, earn points and access exclusive rewards. Sign up now and start earning benefits.
