Different Rules for Different AI Services

Under the government's policy, ChatGPT and Claude are blocked from official use, while Copilot is permitted. The distinction reflects the Swedish government's assessment of the contractual and data-protection conditions attached to each service rather than a blanket rejection of generative AI.

Microsoft's Copilot, offered through enterprise agreements, operates within frameworks that satisfy the government's requirements for handling sensitive information. The chatbot services ChatGPT and Claude, by contrast, did not meet the conditions the government set for official deployment.

Data Protection at the Center of the Decision

The core of the policy lies in data protection. Government agencies handle information that is subject to strict confidentiality obligations, and any AI tool used by public employees must guarantee that input data is not used for model training or exposed to unauthorized access.

Enterprise versions of AI services typically include contractual guarantees on data handling, and the Swedish government's decision indicates that Copilot's terms were judged to provide sufficient safeguards, while the terms for ChatGPT and Claude did not clear the same bar.

Signal for Public-Sector AI Adoption

The Swedish decision offers a reference point for other governments weighing how to introduce generative AI into public administration. Rather than banning AI outright, Sweden has drawn a line based on the commercial terms and data guarantees each vendor offers, allowing compliant services while excluding others.

The move shows that public-sector adoption of AI tools is proceeding selectively, with procurement decisions driven by data-security assessments rather than by the capabilities of the models themselves.

The claim that the Swedish government blocks ChatGPT and Claude while permitting Copilot, with the difference explained by data-handling terms, is True.