← Back to overview

The European Parliament Disables AI on Its Devices – What It Means for Public-Sector Security

The European Parliament has disabled built-in AI features on staff devices. The decision highlights risks of cloud-based AI in critical operations – and points toward local alternatives.

Läs på svenska

The European Parliament has disabled built-in AI features on the smartphones and tablets used by its members and staff. The decision, communicated internally in February 2026, covers writing assistants, text summarization, virtual assistants, and similar features that rely on cloud-based processing.

The reason is straightforward: Parliament’s IT department could not guarantee that sensitive information would not leave devices through the cloud services powering those features. In an environment where draft legislation, internal correspondence, and political deliberations are handled daily, the risk was deemed unacceptable.

Not a Technology Ban – a Risk Decision

The decision does not mean the European Parliament is turning its back on AI. Standard tools like email and calendars are unaffected. What has been disabled are the built-in AI features that automatically process data through external cloud services, often without the user having full visibility into where the information goes.

Parliament also urged members to review AI settings on their personal devices – a signal that the boundary between work and personal devices is blurred in practice.

The decision follows a pattern. As early as 2023, Parliament banned TikTok on staff devices. Discussions are now underway to reduce reliance on non-European productivity tools in favor of European alternatives – part of a broader push toward digital sovereignty.

Relevance for Sweden

For Swedish municipalities, regions, and authorities with critical public-sector mandates, the question is the same. The NIS2 Directive, the AI Act, and the Cyber Resilience Act all require control, traceability, and transparency. Yet many organizations already depend on AI features that are difficult to audit or govern.

The European Parliament’s decision illustrates a paradox: the very institutions driving AI regulation forward are simultaneously reluctant to use the technology internally when security concerns weigh heavily enough.

This does not necessarily mean organizations should disable all AI. But it raises a concrete question: do we know where our data goes when we use the AI tools already enabled in our systems?

Local Alternatives Exist

One viable path forward is to run AI capabilities locally, without cloud dependencies. This means data never leaves the organization’s own infrastructure, providing full control and auditability – qualities that both NIS2 and the AI Act presuppose.

CivSec works with exactly this type of solution for critical public-sector operations: hardened on-premises environments, local generative AI models, and secure development environments built for strict compliance.

Those who want to see how it works in practice will have the opportunity at Mötesplats Samhällssäkerhet, March 17–18 in Stockholm.


Source: The European Parliament pulls back AI from its own devices, The Next Web, February 17, 2026.