Sweden Restricts ChatGPT and Claude for Government Work

Civil servants in Sweden will no longer be able to use OpenAI's ChatGPT or Anthropic's Claude for official work. Citing the risk of sensitive information being entered into unmonitored AI platforms, the government has added both tools to a list of unapproved services. The list also includes DeepSeek, which was banned earlier.

The decision does not end AI use in government. Regulated, approved platforms remain available, chiefly Microsoft Copilot, though it will now come with tighter security controls. A limited number of employees will also keep access to ChatGPT and Claude through a dedicated platform set aside for AI assistants. The restriction targets platforms that fall outside oversight, not AI use as a whole.

Shadow IT and the information security worry

The core concern is tools that employees use without institutional approval, a practice known as shadow IT. Mikael Wettercrantz, Sweden's Chief Information Officer and Director of Digitalisation, said such unmonitored platforms could have consequences for the country's information security.

Earlier this year, a review of the AI tools used by staff at the Swedish Transport Administration found that AI platforms were being used in a "problematic" way for official work.

Why an enterprise agreement matters

Without a proper enterprise agreement and regulation, sensitive information typed into a chatbot for official purposes can flow back to the developer. That data may then be stored and used to train models.

The absence of an enterprise agreement also makes it harder for an organization to monitor how AI tools are being used internally. And when data that should be protected under European rules is processed in, say, a US data center, it leaves Europe's digital borders and can raise additional data protection problems.

Limited access and the DeepSeek precedent

The government said a limited number of employees will continue to reach ChatGPT and Claude through a platform dedicated to AI assistants. Microsoft Copilot remains open to public employees following an internal review, with stricter security controls attached.

This is not the first time Sweden has restricted an AI tool in government work. Like several other countries, it previously banned DeepSeek in the public sector over privacy concerns under the GDPR. The United States banned all DeepSeek products and services over fears that Chinese data laws could give authorities access to the company's servers, and the US Department of Homeland Security warned that the platform could harbor malware and other security vulnerabilities.