The Swedish Government Office will ban ChatGPT and Claude, retaining only Microsoft Copilot

According to Sveriges Radio, the Swedish government office announced that starting from October 26 (Monday), ChatGPT and Claude will be banned on employee computers, with Microsoft Copilot approved as the only AI tool that has passed security review, and civil servants are required to switch to using Copilot.

Mikael Wettercrantz, the agency's Director of Digitalization, stated in a post on the employee intranet that using AI services without a formal contract could "have consequences for our information security." He noted that multiple cases of improper use of open AI services have been identified, involving the sensitivity of the information inputted and the handling of AI-generated content. Due to the lack of formal agreements, consumer-grade AI tools send user queries to external company servers, and U.S. law allows authorities to request these companies to hand over data as required.

Copilot is allowed to remain because the government office enhanced its security settings after review, while ChatGPT and Claude did not receive the same treatment. Some employees can still access these two banned tools through the AI platform built internally by the agency. Previously, the Swedish Tax Agency and the National Board of Health and Welfare had implemented similar bans, and the enforcement agency is also about to make a decision; an internal audit by the Transport Agency found about 10 AI services being "widely and at high risk" used.