On June 3, 2026, the European Commission unveiled the so-called package of technological sovereignty. Alongside the Chips 2.0 Act proposal, the package included the draft of the so-called Cloud and AI Development Act. The regulation establishes a four-level sovereignty framework for cloud infrastructure, reserving the most stringent levels—European ownership and guarantees of non-interference from foreign jurisdictions—for critical services and public administrations.
The proposal has drawn the anticipated negative reaction from the major tech companies. Specifically, they criticized the requirements for levels 3 and 4, namely European ownership and the guarantee of non-interference from foreign jurisdictions. The position of the Silicon Valley giants has lost all footing after the Trump Administration directive limiting access to the Fable 5 and Mythos 5 models exclusively to Americans. But the New York magnate’s decision—made without notice and unilaterally—not only dismantles Silicon Valley’s arguments: it also exposes the limits of the Commission’s own proposal.
The Trump Administration directive is a further stone in the construction of the pyramid of distrust toward technology originating in the U.S. being used by other sectors of the economy. What Farrell and Newman described in 2019 as weaponized interdependence —the coercive use of economic dependence as a tool of power— is no longer a theoretical hypothesis: it is the Trump Administration’s applied technology foreign policy.
“The New York magnate’s decision not only dismantles Silicon Valley’s arguments: it also exposes the limits of the Commission’s own proposal”
First, it sets limits on the technical transparency of American digital goods and services. The veto affects both Fable 5 —a commercial product— and Mythos 5, the model Anthropic had kept restricted since April for its advanced capabilities in identifying cybersecurity vulnerabilities. On June 2, Anthropic provided the language model to governments in fifteen other countries. That access lasted less than two weeks. The window of technological transparency for AI closed before it could be institutionalized.
Second, the Trump Administration’s decision brings to life the ability to trigger the killing switch—the stop button—on any American-originated technological advance. Among the reasons that made the Cloud and AI Development Act possible is the fear of Europe’s excessive dependence on U.S. cloud providers. Specifically, the market share jointly held by Amazon, Microsoft, and Google is estimated at around 80%. From these tech companies, the idea of a prospective “European purchase” policy—driven by fear of a unilateral shutdown of their services by a U.S. government order—had been labeled aggressive and catastrophist. That dystopian scenario has become a reality for European businesses and citizens who have lost access to Fable.
The unilateral shutdown of Anthropic’s models is not without precedent. Huawei’s forced removal from the global tech market in 2019 or the 2022 semiconductor restrictions on China were warnings Europe chose not to heed. We are witnessing the extrapolation to technology of the poem attributed to the Protestant pastor Martin Niemöller, survivor of the Nazi concentration camps: “First they came for Huawei, and I stayed silent because they did not touch Nokia and Ericsson.”
“The joint market share held by Amazon, Microsoft, and Google in this field is estimated at around 80%”
The situation created by the Trump Administration imposes urgency for the EU to adopt a reinforced package of technological sovereignty, at least as much as was pursued to approve the AI Act simplification. The two proposed rules are necessary, but insufficient in both scope and reach. For instance, the sovereignty framework for cloud infrastructure is limited to the application level and fails to address the hardware supply chain. Europe needs to go further in the current geopolitical landscape. The objective must be European strategic autonomy across the entire technology stack, based on both the development of own capabilities and the building of alliances to foster a supply chain resilient to coercion by third parties.
The toolbox for technological sovereignty should be both defensive and offensive. On the defensive front, the most immediate lever is the definition of public procurement requirements and the conditions for receiving state aid across all technological domains. On the offensive front, the EU possesses its own assets that have yet to be deployed as negotiation levers: ASML’s lithography tools —without which no advanced chipmaker can operate— or the telecoms equipment from Ericsson and Nokia are critical links in global value chains.
Stating the same logic as Washington in international tech policy is not a provocation: it is reciprocity. Technological sovereignty is not proclaimed; it is built and defended.
