Washington’s AI insecurity is edging the world towards catastrophe

Lim Teck Ghee

At an online event on 3 September, while touting his agency’s latest report, Superpowers and AGI, CNAS security analyst Jacob Stokes called for the US government to consider diplomatic, espionage, cyber and kinetic measures.

Stokes, deputy director of the Indo-Pacific Security Program at the Centre for a New American Security (CNAS), said the aim was to prevent China from achieving Artificial General Intelligence (AGI) first.

Established in 2007, Western media laud the CNAS for its “pragmatic, data-driven research”. It is seen as an “ideological bridge between administrations, staffed heavily by former high-ranking national security and intelligence officials”.

CNAS reports on the China AI threat to the US are taken extremely seriously by federal policymakers, Congress and the current administration. The think tank is not seen as an outside voice, but as an intellectual architect of current American technology policy.

The gravity with which policymakers treat recent CNAS findings – such as their reports on Red Lines, Adversarial Distillation and Superpowers and AGI – can be seen in a string of recent government actions.

This latest public discourse calls for pre-emptive military or kinetic action (physical force) – such as targeting data centres. It reflects an escalating and dangerous shift in how Washington frames the AI race. It is moving from economic and technical competition to treating AGI as an existential national security risk comparable to nuclear parity.

For now, we can see this AGI/AI dynamic unfolding in the following way.

American sphere: Legislative and policy measures

In the US, the fear of losing technology leadership has resulted in bipartisan consensus favouring containment, reshoring and aggressive export controls.

Export controls and ‘choke-point’ capital restrictions

Hardware and compute blockades: The US Department of Commerce, via the Bureau of Industry and Security, has continually tightened restrictions. These cover the export of advanced AI accelerators (such as high-end Nvidia and AMD chips), semiconductor manufacturing equipment (extreme ultraviolet deep ultraviolet lithography systems) and high-bandwidth memory (HBM).

Outbound investment screening: Executive orders and federal legislative pushes restrict US venture capital and private equity from funding Chinese AI, quantum computing, and semiconductor firms.

Supply chain de-risking: The CHIPS and Science Act channels tens of billions of dollars into domestic fab construction (Intel, TSMC Arizona, Samsung Texas). This aims to remove vulnerabilities and risks in East Asia.

Defence integration and kinetic alignment

National security directives on AI: Federal policies require defence agencies to integrate AI into intelligence, autonomous weapons, and command-and-control frameworks. At the same time, they must ensure critical domestic infrastructure is secured against foreign cyber-threats.

Think tank to war game pipeline: Proposals from major Washington think tanks are frequently integrated into Pentagon scenario planning. Options range from offensive cyber-sabotage to kinetic military contingency planning for high-stakes AGI scenarios. This signals that AI infrastructure is increasingly classified as “critical military targets” in conflict modelling.

Multilateral coalitions and standards

The ‘small yard, high fence’ doctrine: Washington is working with allies – including the EU, Japan, South Korea and the Netherlands – to create aligned export regimes. These are designed to restrict China’s access to global technology supply chains.

Global AI standards: Legally binding framework proposals and international summits aim to establish Western norms for frontier AI deployment, safety and guardrails.

Chinese sphere: Counter-measures

Beijing views US containment actions as an attempt to suppress its economic development and national sovereignty. In response, China has adopted a broad strategic posture centred on self-reliance and asymmetric capabilities.

Full-stack self-reliance (‘indigenisation’)

Native compute acceleration: China is pouring heavy state subsidies into domestic semiconductor design and fabrication (eg Huawei’s Ascend ecosystem and SMIC foundry operations). The aim is to bypass reliance on Western chips.

Algorithmic and architecture optimisation: Faced with hardware constraints, Chinese researchers focus heavily on open-source foundation models, parameter-efficient fine-tuning, and efficient compute architecture. The goal is to squeeze maximum performance out of legacy or non-frontier hardware nodes.

Resource allocation and state infrastructure

‘Eastern data, Western computing’ project: Beijing has centralised its national computing power infrastructure. It is building massive data centre hubs in energy-rich western provinces to maximise efficiency and build energy resilience.

Embodied AI and industrial integration: Recognising potential delays in pure software LLM races, China is leveraging its dominant manufacturing base. This is helping it to lead in ‘embodied AI’ – robotics, industrial automation, autonomous logistics and smart grid systems.

Geopolitical and supply chain counter-measures

Critical mineral control: China leverages its near-monopoly on critical raw materials – gallium, germanium, antimony and rare earths – essential for advanced electronics and military tech. It imposes targeted export restrictions to retaliate against Western sanctions.

Digital Silk Road expansion: Through the Belt and Road Initiative (BRI), China exports its digital infrastructure, cloud computing and telecoms networks across the Global South. This creates an alternative tech ecosystem independent of US platforms.

Defence and hardening tactics

Sovereign cloud and physical hardening: Chinese media and analysts view US talk of targeting data centres as an escalation of threat. In response, critical digital infrastructure is increasingly treated as defence assets – hardened physically, dispersed geographically, and protected with air defence networks and cyber-security protocols.

Catastrophic outcome?

The US strategy for now focuses on containment, compute denial and domestic reshoring to maintain a technological lead.

China’s counter-strategy relies on indigenisation, resource leverage and rapid deployment of AI into real-world manufacturing and robotics.

The inclusion of kinetic strikes on China in mainstream American public policy discourse is telling. It highlights how AI has shifted from a commercial sector into a core domain of strategic geopolitical rivalry.

For now too, it is clear that cooperation and collaboration on AI/AGI to benefit both countries and the rest of the world is an impossible dream.

Meanwhile, the dynamics on the American side have opened the door to several dangerous scenarios: a military option deliberately embarked upon, an over-automated command system triggering runaway escalation, or an automated miscalculation.

Both human miscalculation (pre-emptive strike) and machine miscalculation (autonomous escalation) advance the Doomsday Clock towards a catastrophic outcome for the US and the rest of the world.

Dr Lim Teck Ghee is a Malaysian economic historian and policy analyst.

Support our work by making a donation. Tap to download the QR code below and scan this QR code from Gallery by using TnG e-wallet or most banking apps: