Swedish government offices to ban ChatGPT and Claude

Commenters largely treated the headline as misleading: this is not a rejection of generative AI, but an attempt to stop staff from feeding government material into personal, free, or otherwise uncontracted ChatGPT and Claude accounts. The prevailing view was that a public institution should route prompts through approved services with auditability, confidentiality terms, retention limits, and clear responsibility when sensitive information is mishandled. Several readers noted that even small companies already impose comparable controls, making the real surprise the lack of a prior policy. The thread turned sour over the approved alternative. Many inferred that Microsoft Copilot is the practical winner because Swedish agencies already live inside Microsoft licensing and cloud contracts, not because it is necessarily the strongest assistant. That raises a familiar tradeoff: centralized governance and existing data-protection agreements versus access to the models employees actually find useful. Critics also connected the decision to broader unease over US cloud providers handling sensitive public-sector data, arguing that a sanctioned Microsoft route does not automatically resolve sovereignty or privacy concerns. A few stressed that the scope is narrower than the headline suggests, applying to one government-office agency rather than every Swedish public body.