Sweden: Government Favors Copilot, Blocks ChatGPT and Claude

Starting Monday, October 26, Swedish government offices will block staff from using ChatGPT and Claude on their work computers. Only Microsoft Copilot will remain allowed after a security review. However, Copilot can also run Claude models.

Data Security: Why Stockholm Is Tightening the Rules

This decision is part of a wider trend. Several public agencies are trying to set rules for using artificial intelligence. One example is the removal of DeepSeek from US federal civilian agencies. Italy also blocked ChatGPT in 2023.

In Sweden, public radio reported the measure, which was later confirmed within the government. Mikael Wettercrantz, the director responsible for digital development in government offices, explained on the internal network that using AI services without formal contracts could put information security at risk.

Government offices pointed to problems involving the sensitivity of the data people share and how AI-generated answers are handled.

The concern is about both the terms of use and the AI models themselves. Public employees send their questions to outside companies, and they cannot always control how those companies process the data. A contract, on the other hand, can set out safety guarantees and explain who is responsible for what.

This is the main difference between Copilot and the two banned tools. After reviewing Microsoft's service, the government offices strengthened its security settings before allowing people to use it. However, some employees can still access ChatGPT and Claude through an internal AI platform. The press office did not say how many employees have this access or what rules apply to them.

Other Swedish Public Agencies Are Following Suit

The restrictions do not apply only to government offices. The Swedish Tax Agency and the Institutional Care Council have already introduced similar measures. The Enforcement Authority, which handles tasks such as collecting unpaid debts, is preparing its own decision.

An internal review of the transport administration also found about ten AI services that were widely used and posed risks. These findings show the gap between how quickly people are adopting AI tools and how slowly suitable rules are being put in place.

Other European institutions are taking similar steps. In February, the European Parliament turned off AI features on its work devices, including features linked to ChatGPT, Claude, and Copilot, because the guarantees for protecting data were not strong enough. In the United States, Anthropic lost its appeal against the Pentagon in September, during a period of tension over how its technology is used.

Copilot Is Allowed, but Claude May Still Be Available

However, Stockholm's choice needs a closer look. Microsoft added Claude models to Copilot at the beginning of 2026. In the European Union, an administrator must choose to turn these models on. They are turned off by default.

The Swedish government offices have not said whether their version of Copilot allows people to use these models. So, banning Claude while allowing Copilot may be possible, but the technical difference between the two services might not be as clear as it first appears.

In practice, Stockholm seems to prefer tools that the government can control through its rules of use, rather than banning an entire technology. It remains unclear exactly what guarantees led officials to choose Microsoft.