Flash

Flash·Policy & Regulation·2026-10-25 02:01

Sweden's Government Offices Block ChatGPT and Claude on Work Computers From Oct 26

Sweden's Government Offices will block OpenAI's ChatGPT and Anthropic's Claude on staff work computers from Oct 26, directing employees to approved tools such as Microsoft Copilot, which passed a security review.

Sunny desk with laptop in front of Stockholm government buildings
Sweden's Government Offices blocked ChatGPT and Claude (AI-generated illustration)

What happened

The decision was first reported on Oct 23 by Ekot, the news service of Sweden's public broadcaster Sveriges Radio. The Government Offices (Regeringskansliet) are the administrative machinery behind the prime minister and cabinet; the block covers direct access from staff work computers and takes effect on Monday, Oct 26.

On approved tools, Microsoft Copilot was approved after an internal security review, during which its security settings were tightened; the specific adjustments were not disclosed for security reasons. The block does not cut off all access: some staff will still reach ChatGPT and Claude through a controlled platform built in-house by the Government Offices. The offices had previously imposed a similar restriction on DeepSeek.

The same trend has spread to other agencies: the Swedish Tax Agency (Skatteverket) and the National Board of Institutional Care (SiS) have imposed similar restrictions, and the Enforcement Authority (Kronofogden) is considering a decision. An internal audit this year at the Swedish Transport Administration found staff using about 10 open AI services, with cases including sensitive information entered into the systems and improper handling of generated answers. Government Offices digitalization director Mikael Wettercrantz wrote on the intranet that using AI services without an agreement could affect information security.

Key facts

  1. Effective date and scope: from Monday, Oct 26, 2026, staff at Sweden's Government Offices cannot directly access ChatGPT or Claude from work computers; the blocked tools are OpenAI's ChatGPT and Anthropic's Claude.
  2. Approved tool: Microsoft Copilot is the approved tool, cleared after an internal security review during which its security settings were tightened; employees are to use pre-approved AI tools.
  3. Controlled exception: some staff will still reach ChatGPT and Claude through a platform built in-house by the Government Offices, with the access route under its control.
  4. Spread across agencies: the Swedish Tax Agency and the National Board of Institutional Care have imposed similar restrictions, the Enforcement Authority is considering a decision, and the Government Offices had previously restricted DeepSeek.
  5. Audit background: an internal audit this year at the Swedish Transport Administration found staff using about 10 open AI services, with problem cases including sensitive information entered and improper handling of generated answers; digitalization director Wettercrantz said using AI services without an agreement could affect information security.

Why it matters

the block targets not the models themselves but direct access without an agreement — approved tools and a controlled platform become the channel for government use of frontier models, with procurement and data terms as the entry bar.
Useful Tap if this story helped you

SourcesSveriges Radio (Ekot, Oct 23, 2026, as reported by CryptoBriefing, https://cryptobriefing.com/sweden-government-blocks-chatgpt-claude-copilot/); AzerNEWS (Oct 23, 2026, https://www.azernews.az/region/266078.html). This article is compiled from public information and does not constitute investment advice.

From HIMIAO Horizon · himiao.ai

Comments

  1. Loading comments…

Related

  1. Oct 18GitHub Copilot to Remove Three Flagship Models on November 16: Claude Opus 4.8/5 and GPT-5.6 Terra Retired
  2. Oct 8USA Today Sues OpenAI: $250M+ Damages Sought, "Hundreds of Thousands" of Articles Allegedly Taken for Training
  3. Oct 8Google Launches Universal Work Agent Gemini Agent, Running Gemini and Anthropic's Claude
  4. Oct 19Replit Switches to OpenAI: ChatGPT Replaces Claude as Default Model
  5. Oct 9UK's ICO Names 10 AI Labs in Training-Data Privacy Crackdown: "No Justification" for Non-Compliance

Today

Oct 25 Sunday

Latest news

All →
  1. Oct 25AWS Open-Sources Infinity: Agents Suspend While Waiting, So Waiting Costs No Compute
  2. Oct 241Password Launches Agents Platform: Agent Logins Without Handling Passwords
  3. Oct 24NVIDIA Releases Nemotron 3.5 Bolt E4B: Local Agents on 4GB of VRAM
  4. Oct 24iFLYTEK Launches Agent OS and Xingchen TaaS: Agents Move From Model Calls to Business Loops
  5. Oct 24Moonshot Releases Kimi K3.1 Pro: 3-Trillion-Parameter-Class Open-Weight Model, Deliberately Curbed Cyber Capabilities
  6. Oct 24Block Unveils Dessa, a Voice Model That Learned to Talk from Speech Alone
Ask the cat