Swedish Government Offices Move to Ban ChatGPT and Claude

When you type a query into an open AI chatbot on your work laptop, you rarely think about where those keystrokes actually end up. For civil servants across Sweden, that convenience is about to hit a brick wall. According to reporting by Sveriges Radio's news desk Ekot published on October 23, 2026, the Government Offices of Sweden (Regeringskansliet) are implementing a hard network block on prominent artificial intelligence tools ChatGPT and Claude, cutting off access from staff service computers starting Monday.

src="https://image-api.sr.se/v1/static/AA/83/4f267bc3-0129-4628-bd75-1e593f208dfc.jpg?w=1024&h=576&q=75" alt="Chat GPT app on a mobile phone" class="w-full h-auto rounded-2xl shadow-lg border border-black/10 dark:border-white/10 object-contain max-h-[550px] mx-auto block" loading="lazy" /The ChatGPT app running on a mobile device. Photo: Kiichiro Sato/AP

Blocking Open AI Tools to Force Approved Alternatives

The decision to sever direct web access to OpenAI's ChatGPT and Anthropic's Claude stems from an internal push to ensure employees stick strictly to government-approved software agreements. Mikael Wettercrantz, Director of Digitalization for the Government Offices, declined an interview with Ekot but shared a statement on the agency intranet emphasizing that utilizing unauthorized AI tools can carry direct ”consequences for our information security.”

Instead of the blocked consumer platforms, staff are being guided toward sanctioned services. Primary among them is Microsoft's AI tool Copilot. Internal documents reviewed by Ekot indicate that the administration has simultaneously upgraded security parameters within Copilot, though the press office declined to detail those technical adjustments due to internal security protocols. Additionally, a segment of personnel will maintain access to an internally developed platform designed to interface safely with ChatGPT and Claude.

Widespread Scrutiny Across Swedish Public Agencies

The Government Offices are far from isolated in their crackdown. Ekot's mapping of public administration tech policies reveals that similar network blocks have already been rolled out at the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (Sis). Meanwhile, the Swedish Enforcement Authority (Kronofogden) confirmed it is actively reviewing its own stance on external language models.

Internal reviews at other major bodies have laid bare the extent of casual, high-risk data sharing by public employees. At the Swedish Transport Administration (Trafikverket), an internal audit conducted earlier this year flagged widespread and problematic usage spanning roughly ten different open AI systems. The audit explicitly cited concerns over both the classification sensitivity of information typed into public prompts and the unverified reliability of the generated output.

src="https://image-api.sr.se/v1/static/AA/83/93bcf280-cf14-475b-8b06-7f9259abfd00.jpg?w=80&h=80&q=75" alt="Trafikverket sign" class="w-full h-auto rounded-2xl shadow-lg border border-black/10 dark:border-white/10 object-contain max-h-[550px] mx-auto block" loading="lazy" /Internal audits at Trafikverket exposed widespread unapproved AI usage.

Managing Shadow AI and Vendor Restrictions

The systemic tightening arrives on the heels of previous administrative clampdowns. Regeringskansliet had already restricted usage of the Chinese AI tool Deepseek earlier this year. Public sector IT departments across Stockholm are currently wrestling with shadow AI—the habit of employees pasting internal documents, unredacted citizen data, and operational drafts into web interfaces to draft reports or summarize emails.

While agencies like the Tax Agency faced intense scrutiny over cloud service migration choices earlier this autumn—drawing police pushback and warnings from data protection lawyers regarding unauthorized data exposure—the current wave of blocks represents a tactical shift toward network-level enforcement.

src="https://image-api.sr.se/v1/static/AA/83/6df7891b-0182-4b2d-ac9a-8f9b2d307761.jpg?w=80&h=80&q=75" alt="Caroline Olstedt Carlström and Tax Agency building" class="w-full h-auto rounded-2xl shadow-lg border border-black/10 dark:border-white/10 object-contain max-h-[550px] mx-auto block" loading="lazy" /Data protection experts have repeatedly challenged public agencies over cloud and AI compliance.

As Monday's network blocks go live, civil servants across the Swedish government will find their browser requests to OpenAI and Anthropic intercepted. IT administrators have not announced whether additional consumer-grade generative models will be added to the blocklist as the government standardizes its administrative software stack around pre-vetted enterprise contracts.