Moscow has abruptly softened a controversial draft law on artificial intelligence, abandoning proposals that would have forced developers to train models exclusively on Russian-language data and stripped away requirements that only Russian citizens build so-called “sovereign” AI systems. The climbdown, confirmed by the office of Deputy Prime Minister Dmitry Grigorenko, represents a significant victory for the country’s tech industry, which had warned that data localization and nationality clauses would cripple model quality and isolate the sector from global progress.

The original bill, intended to create a legal foundation for AI in Russia, contained provisions mandating that any model seeking official recognition as national or sovereign be trained solely on data generated by Russian citizens and legal entities. Business groups argued that the volume of high-quality Russian-language material in open sources was far too small to sustain competitive neural networks, risking what one official described as a “degradation of quality.” In the revised text, developers may now use any available data, and foreign specialists can contribute to Russian AI, provided the project is housed in a Russian legal entity that vouches for compliance with domestic law and “traditional spiritual and moral values.” A requirement for platforms with more than half a million users to register as information-dissemination organisers was also struck from the document.

Viewed from St. Petersburg, the episode fits a larger pattern. Yuri Kolotaev, a senior lecturer in European studies at the city’s state university, notes that governments everywhere now treat AI as a strategic resource, much as they once sought to regulate the internet. Any technology that reaches that threshold, he argues, inevitably confronts state-drawn boundaries, even if the tools themselves lack a fixed territorial home. The Russian reversal suggests those boundaries are being drawn with more pragmatism than ideology — a recognition that overly rigid sovereignty requirements risk leaving a nation’s AI ecosystem stranded while rivals advance.

While Moscow debates the geopolitics of data, universities across Latin America are wrestling with a different but related question: how to build critical awareness of an intelligence that is anything but neutral. In Mexico City, Luis Josué Lugo Sánchez of the National Autonomous University’s Centre for Interdisciplinary Research has issued a stark warning that AI is already firmly embedded in lecture halls and laboratories, quietly reconfiguring academic authority. The technologies, he told a recent conference, are shaped by concentrated economic power and respond to political interests, demanding a careful ethical and pedagogical response rather than passive acceptance.

Analysts in Buenos Aires observe a parallel effort to arm citizens with practical skills and scepticism. The University of Buenos Aires has opened enrolment for a suite of free, hybrid-format courses — covering everything from prompt design to predictive analysis — with the explicit goal of nurturing “critical users” equipped to understand the algorithmic forces now shaping the workplace and the market. A second phase aimed at small and medium-sized enterprises is already on the drawing board, reflecting an awareness that AI literacy is fast becoming a baseline for economic competitiveness rather than an academic indulgence.

Taken together, these developments illuminate the central tension of AI governance in the mid-2020s. States are simultaneously competing to attract AI development, defending their cultural and political turf, and belatedly realising that citizens need a new kind of fluency to navigate an algorithmically saturated world. Moscow’s regulatory retreat reveals the limits of sovereignty in a data-driven age; the Mexican and Argentine cases show that the classroom has become the frontline where societies are deciding whether they will be shaped by artificial intelligence or learn to shape it themselves.