Insights - Atlas AI Institute

As artificial intelligence becomes strategically important, governments are increasingly discussing “sovereign AI”—the ability of countries to maintain meaningful control over the AI capabilities, infrastructure, data, talent and institutions that matter to their national interests. But sovereignty does not necessarily mean building a domestic frontier model or becoming technologically self-sufficient. The emerging debate suggests a more practical interpretation: countries need sufficient capacity to access, understand, evaluate, adapt and govern AI without becoming completely dependent on external systems. This article examines the meaning of sovereign AI, why it is emerging, how Europe and India are approaching the issue, and what AI sovereignty could realistically mean for developing countries such as Bangladesh.

AI Governance, Insights

Compute Is Governance

Atlas AI Institute

Artificial intelligence policy is increasingly becoming infrastructure policy. The ability to develop, evaluate and deploy advanced AI depends on access to computing power, specialised chips, cloud infrastructure, data centres, energy, networking and technical talent. As governments invest in AI infrastructure and attempt to secure access to advanced computing, compute is becoming more than a technical resource: it is becoming a strategic component of AI governance. This article examines why compute matters for regulation, national AI capacity, technological sovereignty, sustainability and the Global South—and why countries that cannot access or understand compute may struggle to implement meaningful AI strategies.

AI Governance, Insights

AI Governance as Geopolitical Competition

Atlas AI Institute

Artificial intelligence is increasingly becoming a geopolitical issue as governments compete not only over advanced models, but also over computing infrastructure, semiconductors, data, talent, standards and the rules governing technological development. The United States, China and the European Union are developing different combinations of innovation policy, national security, technological sovereignty and AI governance. At the same time, countries across the Global South face a different challenge: how to participate in an emerging AI economy without becoming permanently dependent on external infrastructure and rule-making systems. This article examines why AI governance is becoming a form of geopolitical competition and what this means for the future of global AI governance.

AI Governance, Insights

Agentic AI: The Next Major Governance Problem

Atlas AI Institute

The next major challenge for AI governance may not come from models that simply generate better answers, but from AI systems that can plan, use tools and take actions on behalf of humans. Agentic AI changes the governance problem because autonomy introduces questions about permissions, human oversight, cybersecurity, accountability, reversibility and continuous monitoring. Governments and standards organisations are beginning to respond, but governance frameworks for autonomous AI remain relatively early. This article examines why agentic AI represents a new phase of AI governance and what governments, companies, universities and developing countries will need to build to govern increasingly autonomous systems.

AI Governance, Insights

Can Governments Actually Control Frontier AI?

Atlas AI Institute

Frontier AI creates a fundamentally different governance challenge from conventional software. The most capable models are developed by a relatively small number of organisations with access to large-scale compute, specialised talent and advanced infrastructure, while governments are still developing the technical institutions required to evaluate and supervise them. Frontier AI governance is therefore not simply a question of writing stronger laws. It involves access to compute, model security, capability evaluations, deployment controls, incident reporting, independent testing and the ability of public institutions to understand rapidly changing systems. This article examines what governments can control, where their authority is limited, and why the future of frontier AI governance may depend on building technical state capacity as much as on regulation.

AI Governance, Insights

Who Audits AI?

Atlas AI Institute

As AI governance moves from principles and legislation toward implementation, a new institutional question is becoming increasingly important: who verifies that AI systems actually meet governance requirements? AI auditing, conformity assessment, technical evaluation and assurance are emerging as critical components of the AI governance ecosystem. The European Union is developing notified-body and conformity-assessment mechanisms, ISO has established standards for AI management systems and the bodies that audit them, while Singapore is building practical AI testing and assurance infrastructure. This article examines what AI auditing means, why independence matters, where technical evaluation fits, and why the ability to assess AI may become a strategic form of governance capacity.

AI Governance, Insights

AI Governance Is Moving from Principles to Implementation

Atlas AI Institute

The first phase of AI governance was largely about principles: fairness, transparency, safety, accountability, human oversight and human rights. The next phase is different. Governments and institutions are increasingly trying to turn those principles into standards, evaluations, audits, documentation, monitoring, enforcement and institutional capacity. The European Union’s AI Act, NIST’s evaluation work and emerging AI safety and assurance institutions illustrate this transition. The central question of AI governance is therefore changing from what responsible AI should mean to how society can verify that AI systems actually meet governance expectations.

AI Governance, Insights

India, Japan, South Korea & Singapore: Alternative Asian Models

Atlas AI Institute

Global AI governance cannot be understood only through the European, American and Chinese models. India, Japan, South Korea and Singapore are developing important alternative approaches that combine AI innovation with national development, technical capacity, safety, standards and institutional coordination. Their experiences demonstrate that countries do not need to choose between regulation and innovation in a simple binary way. Instead, different governance architectures can emerge around national priorities, institutional capacity and technological conditions. This article examines the four Asian approaches and explores what they reveal about the future of AI governance in the Global South.

AI Governance, Insights

China Model: AI Governance + State Strategy + Standards

Atlas AI Institute

China has developed a distinctive AI governance architecture that combines regulation, cybersecurity and data governance with industrial strategy, technical standards and state-led coordination. Rather than separating AI regulation from national technology policy, China’s framework connects the governance of algorithms and generative AI with broader objectives involving security, innovation, infrastructure, data and technological capability. This article examines the evolution of China’s AI governance model, its regulatory instruments, standards ecosystem and strategic characteristics, while considering what the Chinese experience reveals about the increasingly geopolitical nature of AI governance.

AI Governance, Insights

US Model: Innovation + Security + Sectoral/State Governance

Atlas AI Institute

The United States has developed a distinctive AI governance model that differs substantially from the European Union’s comprehensive regulatory approach. Rather than relying primarily on one horizontal AI law, the US system combines presidential policy, federal agencies, technical standards, sector-specific regulation, national-security measures, infrastructure policy and increasingly significant state-level rules. This article examines how the American model developed, why NIST became central to AI governance, how national security and industrial policy have become intertwined with AI regulation, and what the US approach reveals about the emerging global competition over AI governance.