Following an order from the United States Government under its export control regulations, on June 12, 2026, Anthropic deactivated its Fable 5 and Mythos 5 models for all non-U.S. users, including the company’s own employees who do not hold U.S. nationality.
Washington justified the measure on national security grounds, citing the detection of a potential method to bypass the system’s safety filters and access advanced cybersecurity capabilities in the Mythos model.
The episode exposes a deep fracture: access to artificial intelligence can change overnight depending on political decisions, unevenly affecting users, companies, and countries, and even within the same organization depending on an individual’s nationality or the regulatory framework under which they operate.
In practice, this means disruptions to ongoing work, projects in development, and systems already integrated into daily operations, as well as an unequal distribution of capabilities across users, companies, and countries.
The world is fragmenting into technological blocs
The Anthropic blocking episode has reopened a debate that is no longer regional but global: who controls access to artificial intelligence and under what rules.
In Europe, the response is structured around the European Technological Sovereignty Package, which includes the forthcoming Cloud and AI Act (CADA), along with initiatives such as the Chips Act 2.0 and programs supporting AI infrastructure and open source. The goal is to reduce dependence on external providers, although the industrial gap with the United States remains significant.
China has opted for a different strategy. The so-called “DeepSeek moment” of 2025 demonstrated that it was possible to develop competitive models with far lower investment than the major US tech giants. Since then, Beijing has accelerated its push for open models, domestically developed chips, and the massive integration of AI into the productive economy. According to recent data, 34% of job functions in Chinese companies already use artificial intelligence tools, the highest corporate deployment recorded to date. China is not trying to win the same race as the United States. It is running a different one. In
Latin America, the picture is more active than is often perceived, although it remains heterogeneous. The region accounts for 14% of global visits to generative AI applications and ranks third worldwide in downloads of such tools, reflecting strong adoption. However, a structural gap persists: it represents 6.6% of global GDP but only 1.12% of global AI investment, and nearly 90% of supercomputing capacity is concentrated in Brazil.
In terms of governance, the Latin American AI Index (ILIA), promoted by ECLAC and Chile’s National AI Center, has consolidated itself as the region’s main measurement tool, while between 2025 and 2026 more than 200 AI-related bills have been introduced, without an integrated regulatory framework yet in place. A region that adopts quickly, but still structurally depends on external technology.
In the Middle East, AI has become a project of economic transformation on a historic scale. The United Arab Emirates, Saudi Arabia, and Qatar have collectively committed around $2.5 trillion to building a third global pole alongside the United States and China. Key initiatives include Humain in Saudi Arabia and Stargate in the UAE, with partnerships with major technology companies and a strong deployment of infrastructure. All of this is supported by abundant energy and sovereign capital.
Overall, the global map is fragmenting into three dynamics: blocs seeking technological autonomy, regions attempting to build their own governance frameworks, and others directly dependent on external infrastructure. The result is a transition toward an artificial intelligence that no longer operates as a unified global system, but as strategic infrastructure subject to political, economic, and territorial control.
A scenario in which access to the most advanced capabilities depends less and less on innovation and increasingly on the geopolitical position of each actor. And where decisions that today seem technical — which model, which cloud, which chip — are, in practice, decisions about sovereignty.
