Key Takeaways
- An ifo Institute survey of 5,600 German firms found that 88% use US cloud, software, or AI services, with 31% reporting strong dependence on American providers [^4].
- Only 32% of firms view their dependence as very risky, and 42% of those aware of the risks plan no counter-measures.
- The June 2026 Anthropic AI shutdown, ordered by the Trump administration, exposed Europe’s vulnerability to unilateral US export controls on frontier AI models.
- Dependence is highest in data-intensive sectors like advertising (70%) and lowest in physical-goods industries.
- Experts argue that individual corporate actions are insufficient and call for coordinated European investment in compute capacity, energy infrastructure, and public procurement.
Introduction
The sudden shutdown of Anthropic’s most advanced AI models in June 2026 sent shockwaves across European industry and government, exposing a strategic vulnerability that had long been acknowledged but rarely felt so acutely. When the Trump administration ordered Anthropic to suspend access to its Fable 5 and Mythos 5 models for all non-US nationals on June 12, 2026, the move demonstrated how quickly critical technological dependencies could be severed by a single geopolitical decision[^1][^3][^6].
Now, a comprehensive survey by the ifo Institute, conducted in July 2026 with 5,600 German firms, quantifies the depth of this reliance and reveals a troubling gap between awareness and action. While the overwhelming majority of German businesses depend on US digital infrastructure, most do not perceive the risk as urgent—and many have no plans to mitigate it. The findings have intensified calls for coordinated European investment to reduce systemic vulnerability [^4].
The Anthropic Shutdown: A Geopolitical Wake-Up Call
On June 12, 2026, US Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei ordering the company to suspend export of its Claude Mythos 5 and Claude Fable 5 models and prevent their use by foreign nationals, whether residing in the US or abroad[^6]. The administration cited concerns that the models could be deployed by intelligence services in countries such as China and Russia and used to commit cyberattacks. According to the government, a safeguard designed to prevent Fable from identifying software vulnerabilities could be bypassed, or “jailbroken,” by malicious actors[^6].
Anthropic stated that the only way to comply with the export controls was to shut down the models entirely, as the architectural impossibility of dynamically filtering restricted jurisdictions within a centralized cloud infrastructure forced a global pullback[^9]. The company disagreed with the government’s assessment, claiming the bypass found only “minor” security flaws that other publicly available AI models could also identify[^6].
The immediate fallout was severe. As cybersecurity firm ESET noted in a public statement, critical systems across Europe “just… stopped” overnight, with no warning and no migration window[^2]. Politicians from France to the Netherlands amplified calls for AI sovereignty, with French presidential candidate Gabriel Attal likening the shutdown to Iran’s blockade of the Strait of Hormuz[^1].
German Firms’ Heavy Dependence on US Digital Infrastructure
The ifo Institute’s July 2026 survey of 5,600 German firms provides a detailed picture of Europe’s digital dependence. According to the findings, 88% of German companies use US cloud services, software, or AI, and 31% report strong dependence on American providers. This reliance is not evenly distributed across the economy: data-intensive service sectors are particularly exposed, with advertising firms reporting a 70% high-dependence rate, while physical-goods sectors show significantly lower levels of reliance.
This pattern reflects broader European trends. According to the European Commission and European Parliament, the EU relies on non-EU countries for more than 80% of its technology and 70% of its cloud computing[^3]. AI research firm Epoch AI estimates that the US and China together control roughly 90% of global AI computing infrastructure, leaving Europe with limited capacity to train or run AI models independently[^3].
The Anthropic episode made these abstract statistics tangible. As Sandra Wachter, a professor of technology and regulation at the University of Oxford, told Fortune:
“Everybody is aware that we are very dependent on technology from the US… But we have never quite felt what it’s meant to be on the shorter side of the stick—on the side that actually has to face the kill switch”[^3].
Awareness Without Action
Perhaps the most striking finding from the ifo survey is the disconnect between dependence and perceived risk. While nearly nine in ten German firms use US digital services, only 32% view this dependence as very risky. Among firms aware of the risks, 42% plan no counter-measures at all.
Those companies that are taking action tend to favor low-cost, incremental measures, such as switching to European or open-source alternatives. Large-scale infrastructure changes remain rare, reflecting the significant cost and complexity of overhauling established digital ecosystems. This preference for minimal intervention underscores a key challenge: individual corporate decisions, however well-intentioned, are unlikely to meaningfully reduce systemic vulnerability.
The survey’s authors argue that the scale of the problem requires a response beyond what any single company can mount. Coordinated European investment in compute capacity, energy infrastructure, and public procurement is needed to lower systemic risk and build genuine resilience.
Toward Coordinated European Investment
The ifo survey’s authors and policy experts converge on a central conclusion: individual corporate actions cannot secure Europe’s digital resilience. The path forward requires coordinated, large-scale investment across multiple dimensions:
- Compute capacity: Europe must urgently address barriers to data center buildout, including high energy costs, slow planning processes, and fragmented capital markets. Building domestic compute infrastructure would not only support European AI development but also create negotiating leverage with US providers[^1].
- Energy infrastructure: Powering new data centers requires reliable, affordable energy—a persistent challenge given Europe’s energy landscape.
- Public procurement: Leveraging collective purchasing power across EU member states could build shared infrastructure and capabilities that no single country could generate alone, modeled on programs like the Eurofighter Typhoon[^1].
- Middle power coordination: Countries within the EU could engage other middle powers—such as the UK, India, South Korea, Japan, and Canada—to align evaluation regimes, coordinate procurement standards, and invest in each other’s AI ecosystems[^1].
The semiconductor supply chain runs through the Netherlands, Japan, South Korea, and Taiwan. AI evaluation and testing capacity is concentrated in the UK. Many commercially significant AI applications are built on top of American frontier models by developers in Sweden, Canada, and India. This interdependence means that export-controlling American models while relying on allied components, talent, testing infrastructure, revenue, and data is a strategy with a limited shelf life[^1].
Conclusion
The June 2026 Anthropic AI shutdown did not fundamentally change Europe’s position on artificial intelligence. Europe was dependent on American AI models and chips before the directive and remains so today. What the episode did change is the political calculus: by revealing who holds the “kill switch” and how easily it can be pressed, the shutdown transformed an abstract vulnerability into a felt experience.
The ifo survey’s findings—that 88% of German firms rely on US digital services, yet only a third view this as very risky and 42% of those aware plan no counter-measures—suggest that the private sector alone will not close the gap. Without coordinated European investment in compute capacity, energy infrastructure, and public procurement, the continent’s systemic vulnerability will persist. As the survey’s authors conclude, individual actions cannot secure resilience; only a collective, well-resourced European strategy can.
Whether the Anthropic episode proves to be a genuine turning point or merely another rhetorical milestone in the sovereignty debate will depend on whether Europe can translate political momentum into concrete, adequately funded action—and whether it can do so before the next time someone reaches for the switch.
References
[^1]: Afek Shamir (Jun 19, 2026). “What Export Controls on Anthropic’s Most Advanced Models Mean for Europe“. AI Frontiers. Retrieved August 2026.
[^2]: ESET (Jun 13, 2026). “ESET Facebook Post on Anthropic Shutdown“. Facebook. Retrieved August 2026.
[^3]: Fortune (Jun 16, 2026). “Anthropic shutdown ignites calls for sovereign AI across Europe“. Fortune. Retrieved August 2026.
[^4] Clemens Fuest, Jonas Hennrich, Klaus Wohlrabe (2026). “Abhängig, aber gelassen? Wie deutsche Unternehmen ihre Abhängigkeit von US-Digitalanbietern einschätzen“. ifo Institute. Retreived August 2026.
[^6]: Jack Aldane (Jun 16, 2026). “US forces Anthropic to shut down latest AI models, citing national security concerns“. Global Government Forum. Retrieved August 2026.
[^9]: Alias Robotics (Jun 16, 2026). “What Recent Events Reveal About AI Dependency in Cybersecurity“. Alias Robotics. Retrieved August 2026.
