When an AI Model Disappears Overnight: Why Digital Sovereignty Matters More Than Ever
For almost three weeks in June, one of the world's most advanced AI models suddenly became unavailable. The reason was not a technical outage, but an export control decision by the US government. Although access has now been restored, the incident serves as an important reminder for European organisations: access to critical AI technology can depend on political decisions made elsewhere.
What's happening
When Anthropic released its new frontier model Claude Fable 5 in June 2026, it quickly attracted attention for its strong reasoning and coding capabilities. Just a few days after its release, however, the US government issued an export control directive citing national security concerns. The concern was that researchers had demonstrated a method to bypass some of the model's safeguards in cybersecurity-related tasks.
The directive required Anthropic to suspend access for foreign nationals. Because the company had no practical way to verify the nationality of every user in real time, it temporarily disabled access to Fable 5 for everyone worldwide. Nearly three weeks later, after additional safeguards had been implemented and reviewed together with US authorities, the export controls were lifted and global access was restored.
Why this matters
Many organisations today build AI applications assuming that cloud-based frontier models will always be available. The Fable incident demonstrates that this assumption may no longer hold.
AI models are increasingly viewed as strategic technologies, similar to advanced semiconductors. Governments may restrict their availability for national security reasons, export control policies, or geopolitical developments. As AI capabilities continue to improve, such interventions may become more common rather than less.
For European companies, this introduces a new type of operational risk. Even if an organisation has invested considerable effort in integrating a particular model into internal workflows, customer services or software products, access could suddenly change because of regulatory decisions outside its control.
This discussion is therefore closely connected to digital sovereignty. Digital sovereignty does not mean rejecting international technology or trying to become completely independent. Instead, it means ensuring that organisations retain meaningful control over critical digital infrastructure, data, and business processes.
How this impacts you
For many organisations, the practical question is no longer whether to use AI, but how to use it without becoming dependent on a single provider or country.
Fortunately, the technological landscape is evolving rapidly. Open-weight models have made remarkable progress over the past year, and many can now deliver excellent performance for business tasks. These models can be deployed inside a company's own infrastructure, in trusted European cloud environments, or even directly on local devices.
Running AI models locally offers several advantages. Sensitive information remains under the organisation's control. Availability no longer depends entirely on external cloud providers. Regulatory changes have less immediate impact.
The same trend is visible at the edge. Modern laptops, workstations and specialised AI hardware increasingly include dedicated processors capable of running sophisticated language models directly on the device. While local models may not yet match every frontier model in every benchmark, the gap is narrowing quickly, especially for everyday business use cases.
This means organisations no longer need to choose between using the latest AI innovations and maintaining control over their own digital infrastructure. Hybrid approaches are becoming increasingly practical, combining powerful cloud models where appropriate with local models for sensitive or mission-critical tasks.
What to do next
The Fable incident should not discourage organisations from adopting AI. Quite the opposite. It should encourage a more resilient strategy.
When evaluating AI solutions, it is worth asking a few additional questions. How dependent is a business process on a single model provider? Could another model be substituted if necessary? Which applications require the very latest frontier models, and which could run effectively on local or European-hosted alternatives?
Many organisations are also beginning to experiment with multi-model architectures. Instead of relying exclusively on one provider, applications are designed so that different models can be used depending on availability, cost, privacy requirements or performance. This approach not only improves resilience but also makes it easier to benefit from rapid developments across the AI ecosystem.
Digital sovereignty is often discussed as a political objective. In practice, however, it is becoming a business consideration. The organisations that prepare today will be better positioned to adapt tomorrow, regardless of how technology or regulation evolves.
If this topic is relevant for your organization, learn more about our executive AI advisory and hands-on workshops to build internal capabilities.