Swedish civil servants are no longer permitted to access web browser versions of OpenAI's ChatGPT or Anthropic's Claude on their work computers. The policy shift follows official cybersecurity reviews warning that sensitive public data could be improperly processed or retained by external artificial intelligence platforms.
According to reporting by Euronews, the nationwide restriction follows an internal investigation into software usage at the Swedish Transport Administration earlier this year. The audit identified multiple "problematic" instances where staff entered sensitive government materials into public AI portals without formal operational agreements governing data protection and storage protocols.
The Shadow IT Risk in Public Administration
Swedish digital infrastructure authorities emphasized that unmanaged generative AI usage presents significant operational exposure. When employees access consumer-facing AI platforms using work hardware, the lack of enforceable contractual safeguards creates vulnerability around whether proprietary or state data is stored, analyzed, or incorporated into future training sets.
Mikael Wettercrantz, director of digitalisation at the Government Offices of Sweden, warned that unvetted software deployments—frequently classified as shadow IT—threaten state cybersecurity architecture. Without institutional contracts explicitly prohibiting data retention or third-party access, standard consumer terms of service apply by default.
Following the directive, public agencies including the Swedish Tax Agency and the National Board of Institutional Care implemented similar restrictions, as detailed by Swedish public broadcaster Sveriges Radio. Meanwhile, the Swedish Enforcement Authority is currently reviewing its internal software policies to determine if equivalent platform curbs are required.
| Platform / Provider | Current Swedish Government Status | Security & Governance Framework |
|---|---|---|
| OpenAI (ChatGPT Browser) | Blocked on official workstations | Unregulated consumer data pathways; risk of model training exposure |
| Anthropic (Claude Browser) | Blocked on official workstations | Lacks government-level data-protection and storage agreement |
| DeepSeek | Prohibited across state systems | Excluded following systemic national security and data sovereignty evaluations |
| Microsoft Copilot | Approved / Recommended | Cleared internal vetting; governed by formal bilateral security contract |
| Swedish Custom AI Platform | Approved for select roles | Internal specialized platform safeguarding state data governance |
Photographic Documentation
Security Concerns — verified documentary press visual.
Approved Alternatives and Controlled AI Access
The restriction is not a blanket ban on artificial intelligence technology across Swedish civil service operations. Instead, the government is funneling staff toward platforms bound by enterprise data covenants. Euronews' coverage confirms that Microsoft’s Copilot has cleared comprehensive national security screenings and operates under a binding agreement that shields public sector information from model retraining.
Additionally, a restricted tier of personnel may still interact with ChatGPT and Claude, but only via an internally developed specialized platform designed by Swedish authorities. This middleware isolates queries, enforces end-to-end auditability, and prevents external entities from archiving public records.
The policy adjustments align with wider European and global scrutiny over generative AI governance. Both ChatGPT and Claude now join China-based DeepSeek on Sweden’s list of unapproved public-sector tools, matching rigorous data compliance standards enforced across Western regulatory agencies.
Frequently Asked Questions
Why did the Swedish government block ChatGPT and Claude?
Officials determined that using browser versions of these tools without enterprise agreements poses severe security risks. Public agencies discovered staff uploading sensitive information that could potentially be retained, analyzed, or incorporated into model training runs.
Can Swedish public employees still use artificial intelligence tools?
Yes. Employees are encouraged to use approved alternatives such as Microsoft Copilot, which operates under an explicit data security agreement with the Swedish state. A small group of authorized staff can also use ChatGPT and Claude strictly through a sovereign, government-built AI portal.
Which state bodies have enacted these AI restrictions?
The restriction applies to the Government Offices of Sweden, alongside measures adopted by the Swedish Transport Administration, the Swedish Tax Agency, and the National Board of Institutional Care. The Swedish Enforcement Authority is currently assessing similar measures.
