Implications of the AI Diffusion Framework

The Biden-Harris administration unveiled a flurry of policies in their last week in office. The most extraordinary among them is the Framework for Artificial Intelligence (AI) Diffusion. It has many goals: preserving U.S. hegemony in AI technology, balancing innovation and national security, and deterring U.S. adversaries from harvesting the strategic rewards of AI. These goals signal the U.S.’s strategic vision for AI, heralding its transformative potential to advance economic prowess and military dominance in the coming years.

Mechanism of the framework

Compute capacity using advanced AI chips is a key infrastructure necessity in developing advanced AI systems. Leveraging the U.S.’s dominance in chips and the AI supply chain, the framework extends the prevailing export controls to cover the entire gamut of AI technology stack, including AI chips, chip-making tools, and closed AI model weights — the key to a trained AI system’s learning and decision-making abilities.

Countries are placed in three tiers, each subject to different levels of restrictions. The first tier comprises key allies, irreplaceable in the AI supply chain. They are unconstrained in their freedom to import AI technology. The third tier covers key adversaries such as Russia, China, North Korea, and Iran. The framework perpetuates the entire spectrum of export controls and prevents the diffusion of advanced AI systems to these countries. The second tier encompasses the rest of the world, including India. The framework establishes a system of limited access to these countries. Companies from the U.S. can tap into the commercial opportunities of AI in these countries as long as they maintain only a tiny share of their total compute capacity there and safeguard against unauthorised access.

The prevailing export controls have always been restrictive against U.S. adversaries. At first glance, this framework seems to extend some of these restrictions on all but closest allies and partners. Indeed, major allies such as Austria and Israel and other strategic partners such as India are not exempt from the rules. However, the limits and caps imposed are free-handed in satisfying any near-term demand for AI chips from anyone but the countries of concern. Even the export restrictions on model weights are for closed models that far exceed any existing advanced AI systems. So, it will be business as usual in the short term.

The entire system is designed to ensure that if advanced AI systems that are more powerful and capable than the prevailing ones are ever developed, it will happen only in trusted American geographies. The framework will likely succeed in its objectives in the short term and accrue the benefits of cutting-edge AI to the U.S. and its closest allies. But it will influence and reshape the global technological and strategic landscape in the long term.

Through a series of executive actions, the U.S. has mandated the concentration of global AI technological capabilities within its borders and those of allied countries. American companies seeking to establish frontier-scale AI capabilities outside the U.S.— whether motivated by lower costs or more favourable policies — will now face significant barriers. Even among key allies exempt from export controls, this policy sets a concerning precedent of unilateral American executive action that could potentially deny them their strategic priorities in the future.

These allies and American corporations will likely hedge against such possibilities, developing contingency plans to protect their technological interests and maintain competitive advantages. In fact, as AI technology advances in the long term, the likelihood increases that America’s plans may yield unintended consequences. Countries, including U.S. allies, will become more motivated to develop independent supply chains and sovereign AI innovation to circumvent U.S. technological constraints. This could fragment the global AI ecosystem that the U.S. currently dominates.

Eroding goodwill with India

The policy framework disadvantages India, positioning it less favourably than anticipated. By discouraging the development of cutting-edge AI systems beyond U.S. borders, the framework may disincentivise investments and operations of Indian subsidiaries of leading American AI companies. This could create conditions that risk drawing away India’s pool of top AI talent and impeding knowledge transfer and technological innovation within the country. This could be a major point of contention for India, given the implications on its technological and economic interests.

The framework evaluates India solely by its technological capabilities in AI, while ignoring the strategic value of the India-U.S. partnership. It also disregards the substantial social capital built between the two countries through recent technology cooperation, including advances in semiconductor manufacturing collaboration. Instead, it serves as a stark reminder of the U.S.-led technology denial regimes that restricted India’s access to nuclear technologies for three decades following its nuclear tests.

There is a disconnect between America’s export control policies and its strategic partnership with India regarding Indo-Pacific cooperation and shared goals of countering Chinese influence in the region. This misalignment risks undermining the bilateral relationship both nations have carefully cultivated in recent years. It could compel India to hedge against its technological and economic dependence on the U.S.

Ashwin Prasad, Research Analyst in the High-Tech Geopolitics Programme at The Takshashila Institution