Context:

• An AI system was asked whether the U.S. submarine’s sinking of the Iranian frigate IRIS Dena in Sri Lanka’s Exclusive Economic Zone (EEZ) was legal under international law. The AI quickly replied that the action was “not illegal.”
• However, the response did not mention the ongoing legal debate over military activities in an EEZ or the fact that many countries interpret the United Nations Convention on the Law of the Sea (UNCLOS) differently from the U.S. and its allies.
• When the response was questioned, citing India’s position that Article 58 of UNCLOS requires coastal-state consent for foreign military activities in an EEZ, the AI revised its answer.
• It admitted that its earlier conclusion relied mainly on Western naval doctrine and legal scholarship.
• This highlighted a systemic bias in AI systems, reflecting the dominance of Western perspectives in their training data, which can have significant geopolitical implications.

AI and Competing Interpretations of International Maritime Law

• Article 58 of the UNCLOS allows foreign states freedom of navigation, overflight, and other lawful uses of the sea in a country’s EEZ. However, countries interpret this provision differently.
• The U.S. and its allies interpret these freedoms broadly, allowing activities such as military exercises, intelligence gathering, submarine operations, weapons testing, and even combat beyond territorial waters.
• In contrast, India and many Global South countries interpret the provision more narrowly.
• They argue that such freedoms must be directly related to navigation or overflight and that foreign military activities in an EEZ require the consent of the coastal state, respecting the “due regard” principle under Article 58(3).

Overlooked Humanitarian Obligation

• The AI system’s response also ignored a humanitarian issue related to the incident. Article 18 of the Second Geneva Convention requires parties in a conflict to take all possible measures to rescue shipwrecked sailors without delay.
• Reports suggested that the attacking submarine left the scene quickly, leaving rescue operations to the Sri Lankan Navy, which responded to the distress call.
• In the incident, 87 sailors died while 32 were rescued. The AI initially failed to consider this legal and humanitarian obligation.

Structural Bias in AI Systems

• The exchange revealed that AI systems are not neutral interpreters of international law.
• Their responses often reflect the perspectives present in their training data, which are largely dominated by Western legal scholarship and institutional sources.
• As a result, Western interpretations of international law tend to appear as the default position, while views from the Global South are treated as secondary or overlooked.

Implications for Global Discourse

• The case of the IRIS Dena sinking shows that conflicts and legal interpretations in regions like the Indian Ocean are increasingly shaped by algorithmic systems and AI tools used by policymakers and analysts.
• If AI systems consistently prioritise certain interpretations of international law, they can influence global discourse and reinforce existing geopolitical power asymmetries.

India’s Strategic Choices in the Global AI Landscape

• The global AI ecosystem is increasingly becoming bipolar, dominated by U.S. and Chinese technological architectures.
• India faces a strategic choice between adopting the U.S. AI stack, the China AI stack, or building a sovereign Indian AI ecosystem.
• Debate Over the U.S. AI Stack
  • The U.S. AI stack offers advanced chips, cloud infrastructure, models, and platforms to trusted partners, presenting a quick path to AI capability.
  • However, reliance on foreign-controlled infrastructure could limit India’s technological sovereignty, as core computing resources and models would remain externally controlled.
• Pragmatic Approach: Focus on AI Applications
  • Some experts argue that India should focus on deploying AI applications rather than competing in building foundational models.
  • Integrating global AI systems into sectors like healthcare, agriculture, education, and governance could deliver faster economic benefits.
• Concerns About Strategic Dependence
  • Supporters of AI sovereignty warn that dependence on foreign foundational models may create long-term risks.
  • U.S.-developed AI systems often reflect Western data, cultural perspectives, and strategic priorities, which may not align with India’s realities.
• Risk of Digital Colonialism
  • Heavy reliance on external AI infrastructure could lead to digital colonialism, where foreign algorithms control data flows, influence innovation pathways, and shape knowledge production within India’s digital ecosystem.

The Path Toward India’s AI Sovereignty

• Need for Indigenous AI Capabilities - India must move beyond being a consumer of foreign AI systems or merely developing applications on external platforms. It needs to produce its own AI models, datasets, and analytical frameworks to maintain technological independence.
• Building Domestic AI Infrastructure - Achieving this goal requires strategic investment in domestic computing power, indigenous training datasets, secure data infrastructure, and AI models that reflect India’s linguistic diversity and social realities.
• Avoiding Cognitive Dependence - Excessive reliance on foreign AI architectures risks outsourcing not only computation but also knowledge creation and interpretation, allowing external systems to shape India’s narratives and decision-making processes.

AI as a Strategic and Civilisational Challenge

• Artificial intelligence is becoming a global strategic competition. Countries that fail to develop their own AI systems may eventually rely on external models to interpret information and shape policy choices.
• India now faces a crucial decision: remain dependent on foreign digital systems or develop a sovereign AI ecosystem.
• Similar to its achievements in space, nuclear technology, and digital public infrastructure, India must build its own independent AI stack.