Summary
On June 12, 2026, the United States government ordered Anthropic to cut access to its Fable 5 and Mythos 5 models for all non-American nationals, including those on American soil. This unprecedented event, analyzed by the Conseil de l’IA et du numérique (CIANum) in its report titled “Dependency Day,” materialized a risk that had until then remained theoretical: the digital kill switch. It confirms what the Cigref estimates at €265 billion, the cost of Europe’s digital dependency on American players. Against this backdrop, the Digital Resilience Index (DRI), launched in January 2026 at the French Ministry of Economy by a consortium led by the Caisse des Dépôts, RTE, and Docaposte, offers a concrete methodology for measuring and reducing these dependencies. The challenge for European companies: selecting their tools with discernment, by integrating sovereign resilience as a strategic decision criterion on a par with performance.
Artificial intelligence is now an indispensable infrastructure. Former Prime Minister Édouard Philippe put it plainly in his reaction to the ban: “AI is now a critical infrastructure, as essential as electricity or the internet. An infrastructure whose models and computing capacity we do not control is an infrastructure that others can switch off.” This statement captures, in a few words, what the Anthropic affair made tangible. For years, Europe’s digital dependency on American tech giants had been documented, discussed, and criticized, without ever generating a response commensurate with the risk.
On June 12, 2026, the risk moved from the realm of warning to that of operational reality. The challenge is no longer to wait for a regulatory response or a public initiative. It falls to companies, CIOs, and decision-makers to take ownership of their technology choices right now, by stopping the rationalization of tools whose availability depends on a foreign government.
On June 12, 2026, Washington Activated the First Global AI Kill Switch
An Export Directive That Changed the Rules of the Game Overnight
The US government required Anthropic to block access to some of its programs for users who are not American citizens. In an official statement, Anthropic announced that “the US government, invoking its national security powers, issued an export control directive suspending all access to the Fable 5 and Mythos 5 AIs by any foreign national, whether located inside or outside the United States.” The ban even extended to Anthropic employees who are foreign nationals. The decision was issued by Howard Lutnick, Secretary of Commerce, without detailed public justification of the security risks invoked. Anthropic stated that “the Trump administration’s letter did not provide specific details about national security concerns” and that other readily accessible AI models can identify the same security vulnerabilities. By choosing compliance over resistance, Anthropic demonstrated concretely that American law takes precedence over any commercial or international partnership consideration. European companies that had integrated these models into their workflows discovered, without warning, that their production tools had just been deactivated by a foreign government.
What This Decision Reveals About American Power Logic in AI
Specialists point to a key strategic rationale: “The United States wants to prevent other countries like China from exploiting American AI to create software that competes with them,” according to Anastasia Stasenko, co-founder of the French start-up Pleias. The capabilities of the Chinese AI model DeepSeek, released in January 2025, had rattled part of the American industry. The ban therefore fits within a logic of economic and technological warfare in which Europe, lacking sovereign models at scale, finds itself absorbing the effects of a battle between two superpowers. The CIANum, in its “Dependency Day” report, identifies a potentially lasting paradigm shift: the transition from AI perceived as a consumer commodity to an AI of scarcity, controlled and managed by a handful of actors. In parallel, the resurgence of the “existential risk” narrative serves to justify the lockdown of the most advanced models, effectively limiting competition and restricting open innovation. For marketing, legal, medical, and research teams that had relied on Fable 5 and Mythos 5, the lesson is stark: no subscription, no contract, and no history of loyalty provides protection against a unilateral decision by a foreign state.
French Political Reactions, Signs of an Awakening That Is Slow to Become Strategy
The blockage immediately fueled the French presidential pre-campaign ahead of 2027, with several declared candidates reacting on X. Gabriel Attal stated “the AI war has already begun,” warning against “the risk of France’s total vassalization.” Bruno Retailleau declared that “a nation that depends on others for its technology is a nation that can be switched off overnight,” calling on France to treat AI like nuclear capability, “a part of our sovereignty.” Jean-Luc Mélenchon refused to see “France become a digital colony of the USA.” These converging positions carry political weight, but their effect remains limited without operational translation.
The CIANum “Dependency Day” Report: An Analysis That Goes Beyond a Single Event
The CIANum states its conclusion unambiguously: “Regularly invoked in recent months, the kill switch and its impact on European strategic autonomy are no longer a hypothesis. While this concrete risk had already manifested in more targeted and quieter ways in the past, the restrictions imposed on these two cutting-edge models mark a further, significant, and unprecedented step toward the practical generalization of this risk.” This formulation is crucial. It means that the June 12 event is not an isolated incident, but a precedent that normalizes a practice. Recent developments surrounding Anthropic raise a broader question: are European companies building their artificial intelligence revolution on infrastructures they do not control? The honest answer, for anyone who examines the composition of most corporate technology stacks in France, is yes. Data processing, CRM, marketing automation tools, content generation models, behavioral analytics platforms: in each of these categories, the dominance of American players is overwhelming. The decision regarding Anthropic gives a concrete face to what this dominance truly implies in terms of business continuity.
The Threat to Open Source and Accelerated Concentration of the AI Market
The CIANum raises an indirect consequence of the event that concerns the entire AI ecosystem: “The trend working against open source undermines our capacity to independently audit models and access training data, but also to use essential tools to combat, for example, the risks facing minors online.” The national security argument mobilized by Washington effectively legitimizes a strategy of closing off the most advanced models, driven as much by the protection of a technological rent as by the prevention of genuine risks. This dynamic mechanically reduces the possibilities for European actors to access training data, audit algorithmic biases, or develop sovereign applications built on open foundations. Some observers, such as Uljan Sharka, CEO of Domyn, see in this prohibition an opportunity for Europe to develop its own AI solutions, thereby reducing its dependence on American tech giants and stimulating local innovation. This optimistic reading is legitimate, provided that the warning signals translate into concrete decisions to restructure technology stacks.
The Four Axes of Action That the CIANum Recommends for Europe
For Europe, the CIANum identifies four priority workstreams to activate simultaneously. On public policy, French and European responses must crystallize around both supply and demand, to foster scaling and connect private and public ecosystems in support of a genuine European market. On foundational infrastructure, Europe must possess sufficient computing power to develop European alternatives, ensuring that this capacity is not directed exclusively toward American players. On talent, the situation represents a momentum opportunity to attract engineers established outside European borders. On model development, France and Europe must build their own frontier AI research capabilities in order to reduce their vulnerabilities as quickly as possible. These four axes form a coherent roadmap. Their implementation, however, demands a shift in posture among private decision-makers: integrating technological sovereignty as a tool selection criterion, on the same level as total cost of ownership or ease of integration.
The Digital Resilience Index: Turning Awareness Into a Governance Tool
A Concrete Compass for Measuring What Organizations Refused to See
The Digital Resilience Index (DRI) was officially launched on January 26, 2026, at the French Ministry of Economy, during the inaugural Digital Sovereignty Summit. Designed as an “ExCo-compatible” governance tool, it aims to make technological dependencies measurable and subject to arbitration across the entire stack, system by critical system. The DRI measures dependencies across all dimensions of digital operations: software, data, infrastructure, technological assets, internal competencies, governance, and resilience to shocks. The index comprises 20 criteria organized into dependency categories: operational, regulatory, technological, and strategic. This level of granularity is precisely what was missing until now: not a statement of intent on sovereignty, but a quantified diagnostic, comparable across organizations and sectors. The DRI is modeled on the same principle as the B Corp environmental label: it promises to concretely measure an organization’s capacity to manage its critical digital dependencies. For a senior leadership team or a risk committee, having a DRI score makes it possible to assess the organization’s actual exposure and engage in arbitration based on data, not impressions.
An Unprecedented Public-Private Alliance That Gives the Tool Its Credibility
The DRI was announced on July 4, 2025, at the Rencontres Économiques d’Aix-en-Provence, in the presence of Clara Chappaz, Minister Delegate for Artificial Intelligence and Digital Affairs. The first companies to pilot it include RTE and Docaposte, as well as Ouest-France, Caisse des Dépôts, CMA-CGM, MAIF, SNCF, Groupe ADP, and Orange. The DRI is carried by the Digital Resilience Initiative, a non-profit association that will become an international non-profit association (AISBL) in 2026 to accompany its European and international rollout. Olivier Sichel, Director General of the Caisse des Dépôts, serves as honorary president of the association. This composition is a mark of seriousness, bringing together critical infrastructure operators, digital sector players, public institutions, and think tanks. The cost of Europe’s digital dependency is estimated at €265 billion by the Cigref, a figure that served as the catalyst for the initiative and that gives the measure of the economic stakes, beyond any political consideration.
From Diagnosis to Decision: Making the DRI a Reflex in Technology Choices
The DRI aims to produce not a “nice-looking number,” but a collective vision that facilitates arbitration, drawing in legal, risk, and finance stakeholders beyond just the technology function. One of its promises is to avoid the well-known syndrome of frameworks piling on top of existing frameworks. Yann Lechelle is explicit on this point: “There is no obligation to purchase, there is no obligation to implement.” The DRI is embedded within a public-private alliance bringing together the Caisse des Dépôts, the Interministerial Digital Directorate (Dinum), the Cigref, Docaposte, RTE, Numeum, and the Strategic Committee of the “Trusted Digital Solutions” sector. The DRI will take the form of visual radars and strategic reports so that each organization can identify its position and act accordingly. It will be supported by an independent scientific committee, as well as a macroeconomic barometer, the Digital Sovereignty Barometer (DSB), which will measure sector-by-sector technological improvement pathways for the French economy. The practical use of the DRI in calls for tender, application portfolio reviews, or digital transformation audits represents a break with the “it works, so we keep going” logic. It imposes a systematic question: could this supplier be cut off tomorrow by a decision from a foreign state?
The Anthropic affair will not be the last. The logic that led to the ban on Fable 5 and Mythos 5 applies to any tool, platform, and service whose decision-making center is located outside Europe. This is not a new warning: it is a warning that has become concrete and documented, to which the CIANum, the Caisse des Dépôts, and their partners have responded with operational tools. The real question is no longer whether dependency is risky, it is, but how quickly organizations are prepared to move beyond it. Choosing a sovereign tool over a dominant one is rarely the path of least resistance in the short term. It is consistently the choice of long-term resilience, and the distinction between the two has never been clearer than it has been since June 12, 2026.
To accelerate this shift in the marketing and commercial technology space, FranceMartech (francemartech.fr) offers a reference directory of French martech solutions, covering more than 15,000 tools across the entire marketing value chain, from customer relationship management to data management, automation, SEO, and email marketing. Supported by SaaS Advisor, a partner of Netino, the platform enables every team to find a French or European alternative to tools such as HubSpot, Moz, Salesforce, Shopify, or Zoom, with free and open listings available to all national solutions. At a time when digital sovereignty is moving from a political ideal to an operational imperative, FranceMartech represents a concrete starting point for rebuilding a technology stack free from American dependency.
Sources
Conseil de l’IA et du numérique (CIANum), “Dependency Day? Fable 5, Mythos 5: Europe at Its Tipping Point,” June 2026, conseil-ianumerique.fr
Digital Resilience Initiative (aDRI) / Caisse des Dépôts, “Launch of the Digital Resilience Index,” Bercy, January 26, 2026, caissedesdepots.fr
Cigref, estimate of the cost of Europe’s digital dependency: €265 billion, cited in the DRI launch statement, July 2025
Franceinfo, “Artificial Intelligence: Five Questions on the Blocking of Anthropic’s Latest AIs by the Donald Trump Administration,” June 2026
LCP Assemblée Nationale, “AI War, Sovereignty: The Blocking of Anthropic by American Decision Enters the French Presidential Race,” June 2026
Euronews, “Suspension of Anthropic AIs: Europe Speaks Out,” June 2026
RTE, press release, “Launch of an Unprecedented European Initiative: The Digital Resilience Index,” July 4, 2025
