Why AI May Not Be Neutral in International Law

The Exclusive Economic Zone (EEZ) under the United Nations Convention on the Law of the Sea (UNCLOS) extends up to 200 nautical miles from a coastal state's baseline. Within this zone, the coastal state has sovereign rights over natural resources such as fisheries, oil, and gas. However, the legal framework also allows other states certain freedoms, including navigation, overflight, and other internationally lawful uses of the sea, as outlined in Article 58 of UNCLOS. The interpretation of these freedoms has become a major subject of international legal debate.

Two broad interpretations exist regarding military activities in an EEZ. The U.S. and many Western countries interpret Article 58 expansively, arguing that the freedoms include activities such as intelligence gathering, submarine operations, military exercises, and weapons testing, provided they occur beyond territorial waters. According to this view, military operations in another country’s EEZ do not require prior permission from the coastal state.

In contrast, India and several Global South countries interpret the provision more restrictively. They argue that military activities that are not directly related to navigation or overflight require the consent of the coastal state. These countries emphasize the clause in Article 58(3) that requires foreign states to show “due regard” for the rights and laws of the coastal state. This divergence reflects broader geopolitical tensions over maritime governance, particularly in strategically important regions such as the Indian Ocean and the South China Sea.

Artificial Intelligence systems are increasingly used to analyse complex issues such as international law, security policy, and geopolitical events. However, these systems are heavily dependent on the data used during their training process. Since much of the globally available academic literature, legal scholarship, and policy analysis originates from Western institutions, AI models often reflect a Western-centric interpretation of global norms and laws.

This structural bias can shape how AI systems interpret legal frameworks such as UNCLOS. For example, in the case discussed in the article, an AI system initially declared that the sinking of an Iranian warship in an EEZ was “not illegal,” reflecting the dominant Western interpretation of maritime freedoms. However, this response overlooked the competing interpretation held by many countries in the Global South, including India, which emphasises the need for coastal state consent for certain military activities.

Such biases are problematic because policymakers, analysts, and researchers increasingly rely on AI-generated insights for decision-making. If these systems systematically privilege one geopolitical perspective over others, they risk reinforcing existing global power imbalances. Therefore, addressing bias in AI systems requires more diverse training data, representation of non-Western scholarship, and stronger participation from developing countries in shaping the global AI ecosystem.

The global Artificial Intelligence landscape is increasingly dominated by two major technological ecosystems—those led by the United States and China. These ecosystems control critical elements of AI infrastructure, including semiconductor manufacturing, cloud computing platforms, large language models, and massive datasets used to train AI systems. As a result, countries that rely heavily on these external infrastructures may become dependent on the technological frameworks and governance norms established by these powers.

This technological dominance can translate into geopolitical influence. AI systems shape how information is processed, analysed, and interpreted, influencing policy decisions in areas such as security, economics, and diplomacy. For instance, AI models trained primarily on Western data may reproduce Western strategic assumptions about international law, governance, and global politics. Similarly, Chinese AI systems may reflect the governance priorities and information frameworks of the Chinese state.

Consequently, control over AI infrastructure and algorithms can shape the global flow of knowledge and decision-making. Countries lacking indigenous AI capabilities may find their policy frameworks indirectly influenced by external technological systems. This dynamic has led to growing concerns about technological sovereignty and the need for nations to develop independent AI capabilities.

A sovereign AI stack refers to the development of a country’s own technological ecosystem for artificial intelligence, including domestic computing infrastructure, training datasets, foundational models, and applications. For countries like India, building such a system is increasingly seen as crucial for maintaining technological autonomy and protecting national interests in an AI-driven world.

Reliance on foreign AI systems can create several vulnerabilities. If the core computing infrastructure, data pipelines, and algorithms are controlled by external actors, a country’s technological capabilities may be subject to external restrictions or geopolitical pressures. This raises concerns about digital sovereignty, as key decisions regarding data governance, innovation pathways, and technological standards may be influenced by foreign institutions.

Developing a sovereign AI ecosystem also allows India to ensure that its linguistic diversity, social realities, and policy priorities are adequately reflected in AI systems. With hundreds of languages and complex socio-economic conditions, India requires AI models that can effectively operate within its unique context. Building indigenous AI capabilities therefore becomes essential not only for economic development but also for safeguarding cultural representation and strategic autonomy.

India currently faces a strategic debate regarding whether it should prioritize building indigenous AI systems or focus on rapidly adopting existing foreign technologies. Pragmatic advocates argue that competing with major technology hubs such as Silicon Valley in developing frontier AI models may not be feasible in the short term. Instead, they suggest that India should leverage existing global AI engines and concentrate on building innovative applications in sectors such as healthcare, agriculture, education, and governance.

However, supporters of AI sovereignty caution against excessive dependence on foreign technological infrastructures. They argue that reliance on externally controlled models and data ecosystems could lead to “digital colonialism,” where the rules of innovation and knowledge production are shaped by foreign entities. In such a scenario, India’s technological development would remain constrained by external systems that may not reflect its strategic priorities or socio-cultural context.

A balanced approach may therefore be necessary. India can integrate with global AI ecosystems to accelerate innovation while simultaneously investing in domestic capabilities such as data infrastructure, indigenous language models, and high-performance computing systems. This dual strategy would allow India to benefit from global technological advances while maintaining long-term strategic autonomy.

The incident involving the sinking of the Iranian warship IRIS Dena in Sri Lanka’s Exclusive Economic Zone illustrates how modern geopolitical events are increasingly interpreted through digital and algorithmic systems. When the legality of the incident was analysed using an AI system, the model initially concluded that the action was not illegal under international law. However, this response reflected the dominant Western interpretation of UNCLOS rather than acknowledging the alternative legal perspectives held by many countries in the Global South.

This example highlights how AI systems may reproduce existing geopolitical power structures embedded within their training data. Since much of the academic and legal literature used to train AI models originates from Western institutions, these models often prioritize Western interpretations of international norms. Consequently, alternative perspectives—such as those advocated by India, Brazil, or South Africa—may receive less prominence in AI-generated analysis.

The incident also demonstrates the broader implications of AI in shaping global discourse. As policymakers, journalists, and researchers increasingly rely on AI tools for analysis, the biases embedded in these systems can influence how international events are understood. This reinforces the need for more diverse datasets and greater participation from developing countries in shaping global AI architectures.

India’s experience in building Digital Public Infrastructure (DPI) provides a useful model for developing an independent yet globally connected AI ecosystem. Initiatives such as Aadhaar, the Unified Payments Interface (UPI), and the India Stack demonstrate how the country can create scalable technological systems that address domestic needs while also gaining international recognition. These platforms have enabled financial inclusion, digital governance, and innovation in the technology sector.

A similar approach can be adopted for artificial intelligence. India can invest in domestic computing infrastructure, open datasets in Indian languages, and research institutions focused on AI development. Government initiatives such as the IndiaAI Mission aim to strengthen domestic capabilities by supporting startups, developing AI research clusters, and promoting collaboration between academia and industry.

At the same time, India can maintain strategic partnerships with global technology ecosystems to access advanced hardware, research collaboration, and innovation networks. By combining domestic innovation with international cooperation, India can build a resilient AI ecosystem that supports both economic growth and technological sovereignty.