BUENOS AIRES, ARGENTINA — October 31, 2026. Argentine researcher and social theorist Flavia Costa has warned that the rapid development of artificial intelligence for military applications could create risks that conventional regulation alone may be unable to contain. In an interview with journalist Hinde Pomeraniec, published by Infobae on October 25, Costa examined the concentration of technological power, competition between the United States and China, and the emergence of increasingly autonomous AI systems. Her analysis raises a fundamental question about the future of global governance: whether democratic institutions can establish meaningful oversight of technologies whose development is increasingly driven by geopolitical competition and commercial interests.
Costa, a researcher at Argentina’s National Scientific and Technical Research Council (CONICET) and a professor at the University of Buenos Aires, directs TecnocenoLab, an observatory dedicated to examining the relationship between technology, society and digital ecosystems. Her latest book, Vidas artificiales: Del accidente pandémico a las guerras de la IA (Artificial Lives: From the Pandemic Accident to the AI Wars), explores the social transformations associated with generative artificial intelligence. Rather than treating AI exclusively as a technological innovation, Costa examines it as an expanding global infrastructure that influences economic organization, political authority, institutional decision-making and everyday life.
One of her principal concerns involves the development of agentic AI, a category of systems designed to pursue objectives, make intermediate decisions and execute sequences of actions with varying degrees of autonomy. Unlike conventional applications that primarily generate responses to individual requests, agentic systems can interact with digital environments and perform multistep operations. Their growing capabilities have intensified discussions about reliability, cybersecurity and the ability of human operators to maintain effective supervision. Costa argues that military applications could make these challenges particularly dangerous, especially when technological systems operate under circumstances involving uncertainty, limited reaction times and potentially irreversible consequences.
Her warning reflects a broader debate among prominent AI researchers, although there is no scientific consensus that advanced autonomous systems are inherently uncontrollable. Costa refers to concerns expressed by researchers including Yoshua Bengio and Geoffrey Hinton, who have emphasized the importance of studying the risks associated with increasingly capable AI. She also acknowledges the opposing perspective associated with Yann LeCun, who has challenged some of the more alarming predictions concerning artificial intelligence. The disagreement illustrates the complexity of assessing emerging technological capabilities, particularly when theoretical possibilities, laboratory observations and established real-world performance are sometimes discussed without sufficient distinction.
The military dimension introduces additional concerns because artificial intelligence can be incorporated into intelligence analysis, surveillance, targeting support, cyber operations and autonomous weapons systems. These applications differ substantially in their technical design and degree of human supervision, making broad generalizations potentially misleading. Nevertheless, the possibility of delegating increasingly consequential decisions to automated systems has prompted international debates about accountability and the preservation of meaningful human control. Costa questions whether legal restrictions can remain effective when governments perceive technological superiority as essential to national security, creating incentives to accelerate development even while acknowledging the associated risks.
Her geopolitical interpretation places the United States and China at the center of this transformation, while rejecting the idea that global technological competition can be understood entirely through a simple confrontation between two powers. Advanced AI depends on international networks involving semiconductor production, computing infrastructure, research institutions, investment capital and energy resources. These networks create interdependencies that complicate efforts to establish technological autonomy. Costa argues that the resulting international environment is better understood as multipolar, with different actors possessing distinct forms of influence and with technological competition unfolding across overlapping economic and political systems.
Beyond military applications, Costa expresses concern about what she describes as the weakening of institutions in Latin America. Her argument centers on the growing influence of digital platforms over activities previously organized through public institutions, including education, healthcare and administrative services. The accelerated adoption of digital technologies during the COVID-19 pandemic contributed to this transition, making online systems increasingly central to everyday civic participation. For Costa, the question is not whether technological innovation should be rejected, but whether governments, universities and public institutions retain sufficient expertise and authority to shape its implementation according to social needs.
Education provides one illustration of this concern. Costa discusses the growing use of artificial intelligence in learning environments and debates surrounding homeschooling in Argentina and Chile. In her view, the expansion of technology-mediated education raises questions about the responsibilities of the state, the continuity of shared educational standards and the social functions traditionally performed by schools. These concerns do not establish that home education or AI-assisted learning necessarily undermines educational outcomes. They highlight a policy debate involving access, quality, institutional oversight and the appropriate relationship between technological flexibility and public responsibility.
The environmental consequences of AI infrastructure also feature in her analysis. Training and operating advanced models require computing facilities whose electricity demand, cooling requirements and resource consumption can influence local communities and national energy systems. Costa points to discussions about developing data-center infrastructure in Argentina’s Patagonia region as an example of how international technology investment intersects with domestic development policy. Such projects can create economic opportunities, but their broader effects depend on energy availability, environmental safeguards, infrastructure planning and the distribution of benefits and costs.
Despite the seriousness of her concerns, Costa distinguishes professional vigilance from fear. She maintains that researchers must remain capable of examining evidence, identifying opportunities for intervention and evaluating competing technological claims without surrendering to fatalism. That distinction is important because the future of AI is not determined exclusively by technological capability. Institutional arrangements, industrial incentives, public oversight and international agreements can influence how systems are developed and deployed, even when the effectiveness of particular regulatory approaches remains contested.
Her arguments ultimately shift the debate away from the familiar question of whether artificial intelligence is beneficial or harmful. AI encompasses different technologies with different applications, risks and social consequences, making a single judgment inadequate. The more consequential issue concerns who controls its development, which institutions can demand accountability and how societies respond when commercial incentives or national security objectives conflict with broader public interests. These questions become especially urgent in military contexts, where failures of oversight can have consequences extending far beyond the organizations deploying the technology.
Costa’s intervention contributes to an expanding international discussion about technological power and democratic responsibility. Her warning that regulation alone may be insufficient should be understood as an argument for examining the limits of existing governance mechanisms, rather than proof that every form of oversight is destined to fail. The unresolved challenge is whether governments, scientific institutions and technology companies can establish credible safeguards while maintaining international cooperation in an environment shaped by strategic rivalry. As artificial intelligence becomes increasingly embedded in economic and security systems, its future will depend not only on what machines can accomplish, but also on the political and institutional decisions determining how their capabilities are used.
Phoenix24: Beyond the technology, the consequences. / Más allá de la tecnología, las consecuencias.