Human in the Loop

In the corridors of power from Washington to Beijing, a new form of competition is taking shape. It's fought not with missiles or marines, but with machine learning models and neural networks. As artificial intelligence becomes increasingly central to military capabilities, the race to develop, deploy, and control these technologies has become a defining feature of contemporary geopolitics. The stakes are immense: the nations that master military AI may well shape the global balance of power for decades to come.

The New Great Game

The parallels to historical great power competition are striking, but today's contest unfolds across silicon wafers rather than traditional battlefields. The primary protagonists are the United States and China, but the competition extends far beyond these superpowers into research laboratories, corporate boardrooms, and international standards bodies worldwide.

This competition has fundamentally altered how nations approach AI development. Where scientific collaboration once flourished, researchers now find themselves navigating national security imperatives alongside the pursuit of knowledge. The open-source ethos that drove early AI breakthroughs increasingly gives way to classified programmes and export controls.

The transformation reflects explicit policy priorities. China's national AI strategy positions artificial intelligence as essential for national competitiveness and military modernisation. The approach represents more than research priorities—it positions AI as a tool of statecraft and national strength, with significant state investment and coordination across civilian and military applications.

The United States has responded through institutional changes, establishing dedicated AI offices within the Department of Defense and increasing investment in military AI research. However, America's approach differs markedly from China's centralised strategy. Instead of top-down directives, the US relies on its traditional strengths: venture capital, university research, and private sector innovation. This creates a more distributed but arguably less coordinated response to the competitive challenge.

The competition extends beyond technological capabilities to encompass the rules governing AI use. Both nations recognise that controlling AI development means influencing the standards and norms that will govern its deployment. This has created a dynamic where countries racing to build more capable military AI systems simultaneously participate in international forums discussing their regulation.

Recent developments in autonomous weapons systems illustrate this tension. Military AI applications now span from logistics and intelligence analysis to more controversial areas like autonomous target identification. These developments occur as AI systems move from experimental add-ons to central components of military operations, fundamentally altering strategic planning, threat assessment, and crisis management processes.

The geopolitical implications extend beyond bilateral competition. As the Brookings Institution notes, this rivalry is “fueling military innovation” and accelerating the development of AI-enabled weapons systems globally. Nations fear falling behind in what they perceive as a critical technological race, creating pressure to advance military AI capabilities regardless of safety considerations or international cooperation.

The Governance Vacuum

Perhaps nowhere is the impact of geopolitical competition more evident than in the struggle to establish international governance frameworks for military AI. The current landscape represents a dangerous paradox: as AI capabilities advance rapidly, the institutional mechanisms to govern their use lag increasingly behind.

The Carnegie Endowment for International Peace has identified this as a “governance vacuum” that poses significant risks to global security. Traditional arms control mechanisms developed during the Cold War assume weapons systems with predictable, observable characteristics. Nuclear weapons require specific materials and facilities that can be monitored. Chemical weapons leave detectable signatures. But AI weapons systems can be developed using commercial hardware and software, making verification enormously challenging.

This verification challenge is compounded by the dual-use nature of AI technology. The same machine learning techniques that power recommendation engines can guide autonomous weapons. The neural networks enabling medical diagnosis can also enhance target recognition. This blurring of civilian and military applications makes traditional export controls and technology transfer restrictions increasingly ineffective.

The institutional landscape reflects this complexity. Rather than a single governing body, AI governance has evolved into what researchers term a “regime complex”—a fragmented ecosystem of overlapping institutions, initiatives, and informal arrangements. The United Nations Convention on Certain Conventional Weapons discusses lethal autonomous weapons systems, while the OECD develops AI principles for civilian applications. NATO explores AI integration, and the EU crafts comprehensive AI legislation.

Each forum reflects different priorities and power structures. The UN process, while inclusive, moves slowly and often produces minimal agreements. The OECD represents developed economies but lacks enforcement mechanisms. Regional organisations like NATO or the EU can move more quickly but exclude key players like China and Russia.

This fragmentation creates opportunities for forum shopping, where nations pursue their preferred venues for different aspects of AI governance. The United States might favour NATO discussions on military AI while supporting OECD principles for civilian applications. China participates in UN processes while developing bilateral arrangements with countries along its Belt and Road Initiative.

The result is a patchwork of overlapping but incomplete governance mechanisms. Some aspects of AI development receive significant attention—algorithmic bias in civilian applications, for instance—while others, particularly military uses, remain largely unregulated. This uneven coverage creates both gaps and conflicts in the emerging governance landscape.

The European Union has attempted to address this through its AI Act, which includes provisions for high-risk applications while primarily focusing on civilian uses. However, the EU's approach reflects particular values and regulatory philosophies that may not translate easily to other contexts. The emphasis on fundamental rights and human oversight, while important, may prove difficult to implement in military contexts where speed and decisiveness are paramount.

Military Integration and Strategic Doctrine

The integration of AI into military doctrine represents one of the most significant shifts in warfare since the advent of nuclear weapons. Unlike previous military technologies, AI doesn't simply provide new capabilities; it fundamentally alters how militaries think, plan, and respond to threats.

Research from Harvard's Belfer Center highlights how this transformation is most evident in what scholars call “militarised bargaining”—the use of military capabilities to achieve political objectives without necessarily engaging in combat. AI systems now participate directly in this process, analysing adversary behaviour, predicting responses to various actions, and recommending strategies for achieving desired outcomes.

The implications extend far beyond traditional battlefield applications. AI systems increasingly support strategic planning, helping military leaders understand complex scenarios and anticipate consequences of various actions. They assist in crisis management, processing vast amounts of information to provide decision-makers with real-time assessments of evolving situations. They even participate in diplomatic signalling, as nations use demonstrations of AI capabilities to communicate resolve or deter adversaries.

This integration creates new forms of strategic interaction. When AI systems help interpret adversary intentions, their accuracy—or lack thereof—can significantly impact crisis stability. If an AI system misinterprets routine military exercises as preparation for attack, it might recommend responses that escalate rather than defuse tensions. Conversely, if it fails to detect genuine preparations for aggression, it might counsel restraint when deterrent action is needed.

The speed of AI decision-making compounds these challenges. Traditional diplomatic and military processes assume time for consultation, deliberation, and measured response. But AI systems can process information and recommend actions in milliseconds, potentially compressing decision timelines to the point where human oversight becomes difficult or impossible.

The challenge of maintaining human control over AI-enabled weapons systems illustrates these concerns. Current international humanitarian law requires that weapons be under meaningful human control, but defining “meaningful” in the context of AI systems proves remarkably difficult. Questions arise about what constitutes sufficient control when humans authorise AI systems to engage targets within certain parameters, particularly when the system encounters situations not anticipated by its programmers.

These questions become more pressing as AI systems demonstrate broader capabilities and greater autonomy. Early military AI applications focused on relatively narrow tasks—image recognition, pattern analysis, or route optimisation. But newer systems demonstrate broader capabilities, able to adapt to novel situations and make complex judgements that previously required human intelligence.

The military services are responding by developing new doctrines and training programmes that account for AI capabilities. Personnel now train alongside AI systems that can process sensor data faster than any human. Commanders work with AI assistants that can track multiple contacts simultaneously. Forces experiment with AI-enabled logistics systems that anticipate supply needs before human planners recognise them.

This human-machine collaboration requires new skills and mindsets. Military personnel must learn not just how to use AI tools, but how to work effectively with AI partners. They need to understand the systems' capabilities and limitations, recognise when human judgement should override AI recommendations, and maintain situational awareness even when AI systems handle routine tasks.

The Innovation-Safety Tension

The relationship between innovation and safety in military AI development reveals one of the most troubling aspects of current geopolitical competition. As nations race to develop more capable AI systems, the pressure to deploy new technologies quickly often conflicts with the careful testing and evaluation needed to ensure they operate safely and reliably.

This tension manifests differently across various military applications. In logistics and support functions, the risks of AI failure might be manageable—a supply prediction error could cause inconvenience but rarely catastrophe. But as AI systems assume more critical roles, particularly in weapons systems and strategic decision-making, the consequences of failure become potentially catastrophic.

The competitive dynamic exacerbates these risks. When nations believe their adversaries are rapidly advancing their AI capabilities, the temptation to rush development and deployment becomes almost irresistible. The fear of falling behind can override normal safety protocols and testing procedures, creating what researchers term a “safety deficit” in military AI development.

This problem is compounded by the secrecy surrounding military AI programmes. While civilian AI development benefits from open research, peer review, and collaborative debugging, military AI often develops behind classified walls. This secrecy limits the number of experts who can review systems for potential flaws and reduces the feedback loops that help identify and correct problems.

The commercial origins of much AI technology create additional complications. Military AI systems often build on civilian foundations—commercial machine learning frameworks, open-source libraries, and cloud computing platforms. But the transition from civilian to military applications introduces new requirements and constraints that may not be fully understood or properly addressed.

The challenge of adversarial attacks on AI systems illustrates these concerns. Researchers have demonstrated that carefully crafted inputs can fool AI systems into making incorrect classifications—causing an image recognition system to misidentify objects, for instance. In civilian applications, such failures might cause inconvenience. In military applications, they could prove lethal.

The development of robust defences against such attacks requires extensive testing and validation, but this process takes time that competitive pressures may not allow. Military organisations face difficult choices between deploying potentially vulnerable systems quickly or taking the time needed to ensure their robustness.

International cooperation could help address these challenges, but geopolitical competition makes such cooperation difficult. Nations are reluctant to share information about AI safety challenges when doing so might reveal capabilities or vulnerabilities to potential adversaries. The result is a fragmented approach to AI safety, with each nation largely working in isolation.

Some progress has occurred through academic exchanges and professional conferences, where researchers from different countries can share insights about AI safety challenges without directly involving their governments. However, the impact of such exchanges remains limited by the classified nature of much military AI development.

Regional Approaches and Alliance Dynamics

The global landscape of AI governance reflects not just bilateral competition between superpowers, but also the emergence of distinct regional approaches that shape international norms and standards. These regional differences create both opportunities for cooperation and potential sources of friction as different models compete for global influence.

The European approach emphasises fundamental rights, human oversight, and comprehensive regulation. The EU's AI Act represents one of the most ambitious attempts to govern AI development through formal legislation, establishing risk categories and compliance requirements that can extend beyond European borders through regulatory influence. When European companies or markets are involved, EU standards can effectively become global standards.

This regulatory approach reflects deeper European values about technology governance. Where the United States tends to favour market-driven solutions and China emphasises state-directed development, Europe seeks to balance innovation with protection of individual rights and democratic values. The EU's approach to military AI reflects these priorities, emphasising human control and accountability even when such requirements might limit operational effectiveness.

The transatlantic relationship adds complexity to this picture. NATO provides a forum for coordinating AI development among allies, but the organisation must balance American technological leadership with European regulatory preferences. The result is complex negotiations over standards and practices that reflect broader tensions within the alliance about technology governance and strategic autonomy.

NATO has established principles for responsible AI use that emphasise human oversight and ethical considerations, but these principles must be interpreted and implemented by member nations with different legal systems and military doctrines. Maintaining interoperability while respecting national differences requires continuous negotiation and compromise.

Asian allies of the United States face their own unique challenges. Countries like Japan, South Korea, and Australia must balance their security partnerships with America against their economic relationships with China. This creates complex calculations about AI development and deployment that don't map neatly onto alliance structures.

Japan's approach illustrates these tensions. As a close US ally with advanced technological capabilities, Japan participates in various American-led AI initiatives while maintaining its own distinct priorities. Japanese companies have invested heavily in AI research, but these investments must navigate both American export controls and Chinese market opportunities.

The Indo-Pacific region has become a key arena for AI competition and cooperation. The Quad partnership between the United States, Japan, India, and Australia includes significant AI components, while China's Belt and Road Initiative increasingly incorporates AI technologies and standards. These competing initiatives create overlapping but potentially incompatible frameworks for regional AI governance.

India represents a particularly interesting case. As a major power with significant technological capabilities but non-aligned traditions, India's approach to AI governance could significantly influence global norms. The country has developed its own AI strategy that emphasises social benefit and responsible development while maintaining strategic autonomy from both American and Chinese approaches.

The Corporate Dimension

The role of private corporations in military AI development adds layers of complexity that traditional arms control frameworks struggle to address. Unlike previous military technologies that were primarily developed by dedicated defence contractors, AI capabilities often originate in commercial companies with global operations and diverse stakeholder obligations.

This creates unprecedented challenges for governments seeking to control AI development and deployment. Major technology companies possess AI capabilities that rival or exceed those of many national governments. Their decisions about research priorities, technology sharing, and commercial partnerships can significantly impact national security considerations.

The relationship between these companies and their home governments varies considerably across different countries and contexts. American tech companies have historically maintained significant independence from government direction, though national security considerations increasingly influence their operations. Public debates over corporate involvement in military AI projects have highlighted tensions between commercial interests and military applications.

Chinese technology companies operate under different constraints and expectations. China's legal framework requires companies to cooperate with government requests for information and assistance, creating concerns among Western governments about the security implications of Chinese AI technologies. These concerns have led to restrictions on Chinese AI companies in various markets and applications.

European companies face their own unique challenges, operating under the EU's comprehensive regulatory framework while competing globally against American and Chinese rivals. The EU's emphasis on digital sovereignty and strategic autonomy creates pressure for European companies to develop independent AI capabilities, but the global nature of AI development makes complete independence difficult to achieve.

The global nature of AI supply chains complicates efforts to control technology transfer and development. AI systems rely on semiconductors manufactured in various countries, software frameworks developed internationally, and data collected worldwide. This interdependence makes it difficult for any single country to control AI development completely, but it also creates vulnerabilities that can be exploited for strategic advantage.

Recent semiconductor export controls illustrate these dynamics. American restrictions on advanced chip exports to China aim to slow Chinese AI development, but they also disrupt global supply chains and create incentives for countries and companies to develop alternative suppliers. The long-term effectiveness of such controls remains uncertain, as they may accelerate rather than prevent the development of alternative technological ecosystems.

The talent dimension adds another layer of complexity. AI development depends heavily on skilled researchers and engineers, many of whom are internationally mobile. University programmes, corporate research labs, and government initiatives compete globally for the same pool of talent, creating complex webs of collaboration and competition that transcend national boundaries.

Immigration policies increasingly reflect these competitive dynamics. Countries adjust visa programmes and citizenship requirements to attract AI talent while implementing security screening to prevent technology transfer to rivals. The result is a global competition for human capital that mirrors broader geopolitical tensions.

Emerging Technologies and Future Challenges

The current focus on machine learning and neural networks represents just one phase in the evolution of artificial intelligence. Emerging technologies like quantum computing, neuromorphic chips, and brain-computer interfaces promise to transform AI capabilities in ways that could reshape military applications and governance challenges.

Quantum computing represents a potential paradigm shift. While current AI systems rely on classical computing architectures, quantum systems could solve certain problems exponentially faster than any classical computer. The implications for cryptography are well understood—quantum computers could break many current encryption schemes—but the impact on AI development is less clear and potentially more profound.

Quantum machine learning algorithms could enable AI systems to process information and recognise patterns in ways that are impossible with current technology. The timeline for practical quantum computers remains uncertain, but their potential impact on military AI capabilities is driving significant investment from major powers.

The United States has launched a National Quantum Initiative that includes substantial military components, while China has invested heavily in quantum research through its national laboratories and universities. European countries and other allies are developing their own quantum programmes, creating a new dimension of technological competition that overlays existing AI rivalries.

Neuromorphic computing represents another frontier that could transform AI capabilities. These systems mimic the structure and function of biological neural networks, potentially enabling AI systems that are more efficient, adaptable, and robust than current approaches. Military applications could include autonomous systems that operate for extended periods without external support or AI systems that can adapt rapidly to novel situations.

The governance challenges posed by these emerging technologies are daunting. Current international law and arms control frameworks assume weapons systems that can be observed, tested, and verified through traditional means. But quantum-enhanced AI systems or neuromorphic interfaces might operate in ways that are fundamentally opaque to external observers.

The verification problem is particularly acute for quantum systems. The quantum states that enable their computational advantages are extremely fragile and difficult to observe without disturbing. This could make it nearly impossible to verify whether a quantum system is being used for permitted civilian applications or prohibited military ones.

The timeline uncertainty surrounding these technologies creates additional challenges for governance. If quantum computers or neuromorphic systems remain decades away from practical application, current governance frameworks might be adequate. But if breakthroughs occur more rapidly than expected, the international community could face sudden shifts in military capabilities that existing institutions are unprepared to address.

The Path Forward: Navigating Chaos and Control

The future of AI governance will likely emerge from the complex interplay of technological development, geopolitical competition, and institutional innovation. Rather than a single comprehensive framework, the world appears to be moving toward what the Carnegie Endowment describes as a “regime complex”—a fragmented but interconnected system of governance mechanisms that operate across different domains and levels.

This approach has both advantages and disadvantages. On the positive side, it allows different aspects of AI governance to develop at different speeds and through different institutions. Technical standards can evolve through professional organisations, while legal frameworks develop through international treaties. Commercial practices can be shaped by industry initiatives, while military applications are governed by defence partnerships.

The fragmented approach also allows for experimentation and learning. Different regions and institutions can try different approaches to AI governance, creating natural experiments that can inform future developments. The EU's comprehensive regulatory approach, America's market-driven model, and China's state-directed system each offer insights about the possibilities and limitations of different governance strategies.

However, fragmentation also creates risks. Incompatible standards and requirements can hinder international cooperation and create barriers to beneficial AI applications. The lack of comprehensive oversight can create gaps where dangerous developments proceed without adequate scrutiny.

The challenge for policymakers is to promote coherence and coordination within this fragmented landscape without stifling innovation or creating rigid bureaucracies that cannot adapt to rapid technological change. This requires new forms of institutional design that emphasise flexibility, learning, and adaptation rather than comprehensive control.

One promising approach involves the development of what scholars call “adaptive governance” mechanisms. These systems are designed to evolve continuously in response to technological change and new understanding. Rather than establishing fixed rules and procedures, adaptive governance creates processes for ongoing learning, adjustment, and refinement.

The technical nature of AI development also suggests the importance of involving technical experts in governance processes. Traditional diplomatic and legal approaches to arms control may be insufficient for technologies that are fundamentally computational. New forms of expertise and institutional capacity are needed to bridge the gap between technical realities and policy requirements.

International cooperation remains essential despite competitive pressures. Many AI safety challenges are inherently global and cannot be solved by any single country acting alone. The global nature of these challenges suggests the need for cooperation even amid broader geopolitical tensions.

The private sector role suggests the need for new forms of public-private partnership that go beyond traditional government contracting. Companies possess capabilities and expertise that governments need, but they also have global operations and stakeholder obligations that may conflict with narrow national interests. Finding ways to align these different priorities while maintaining appropriate oversight represents a key governance challenge.

The emerging governance landscape will likely feature multiple overlapping initiatives rather than a single comprehensive framework. Professional organisations will develop technical standards, regional bodies will create legal frameworks, military alliances will coordinate operational practices, and international organisations will provide forums for dialogue and cooperation.

Success in this environment will require new skills and approaches from all participants. Policymakers need to understand technical realities while maintaining focus on broader strategic and ethical considerations. Technical experts need to engage with policy processes while maintaining scientific integrity. Military leaders need to integrate new capabilities while preserving human oversight and accountability.

The stakes of getting this right are enormous. AI technologies have the potential to enhance human welfare and security, but they also pose unprecedented risks if developed and deployed irresponsibly. The geopolitical competition that currently drives much AI development creates both opportunities and dangers that will shape the international system for decades to come.

The path forward requires acknowledging both the competitive realities that drive current AI development and the cooperative imperatives that safety and governance demand. This balance will not be easy to achieve, but the alternative—an unconstrained AI arms race without adequate safety measures or governance frameworks—poses far greater risks.

The next decade will be crucial in determining whether humanity can harness the benefits of AI while managing its risks. The choices made by governments, companies, and international organisations today will determine whether AI becomes a tool for human flourishing or a source of instability and conflict. The outcome remains uncertain, but the urgency of addressing these challenges has never been clearer.

References and Further Information

Brookings Institution. “The global AI race: Will US innovation lead or lag?” Available at: www.brookings.edu

Belfer Center for Science and International Affairs, Harvard Kennedy School. “AI and Geopolitics: Global Governance for Militarized Bargaining.” Available at: www.belfercenter.org

Carnegie Endowment for International Peace. “Governing Military AI Amid a Geopolitical Minefield.” Available at: carnegieendowment.org

Carnegie Endowment for International Peace. “Envisioning a Global Regime Complex to Govern Artificial Intelligence.” Available at: carnegieendowment.org

Social Science Research Network. “Artificial Intelligence and Global Power Dynamics: Geopolitical Implications and Strategic Considerations.” Available at: papers.ssrn.com

Additional recommended reading includes reports from the Center for Strategic and International Studies, the International Institute for Strategic Studies, and the Stockholm International Peace Research Institute on military AI development and governance challenges.

---

Tim Green UK-based Systems Theorist & Independent Technology Writer

Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.

His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.

ORCID: 0000-0002-0156-9795 Email: tim@smarterarticles.co.uk

Discuss...