ENISA, the European Union's cybersecurity agency, sat down with Anthropic, the AI safety company behind Claude, this week. The discussion took place against the backdrop of tightening US export controls on advanced AI models — a policy that underscores just how vulnerable the EU remains when it comes to critical AI infrastructure. The meeting, first reported by Crypto Briefing, adds a new layer to the bloc's long-running conversation about technological sovereignty.
Why the meeting matters
ENISA's job is to keep Europe's digital infrastructure secure. Anthropic builds large language models that are increasingly embedded in everything from customer service bots to code generators. If the US restricts exports of those models — or the chips needed to train them — European companies and governments could be left scrambling for alternatives. A direct dialogue between the cybersecurity regulator and a leading AI lab suggests Brussels is already trying to get ahead of that risk.
The export control context
The Biden administration, and now the Trump administration, has steadily tightened the rules on sending advanced AI chips and model weights abroad. The logic is national security: keep cutting-edge AI out of rivals' hands. But the policy doesn't distinguish between adversaries and allies. European firms have complained that they're effectively being cut off from the technology they need to compete. The ENISA-Anthropic meeting is a signal that the EU is taking the threat seriously — and looking for ways to secure access or build its own capability.
Sovereignty debate heats up
This isn't the first time the EU has talked about strategic autonomy in tech. But the export controls make the conversation more urgent. If Europe can't reliably import the best AI models, it needs to develop them at home. That means more funding, more compute, and more cooperation between regulators and the private sector. The meeting with Anthropic — an American company — is notable precisely because it shows the EU is still working with US firms, even as it tries to reduce dependency. Whether that balancing act can last is an open question.
The next concrete step: the European Commission is expected to release a white paper on AI compute infrastructure later this summer. How much of that plan relies on homegrown models — and how much on continued access to US technology — will be the real test.
