What is the New Rule, Exactly?
On November 16, 2026, the U.S. Commerce Department's Bureau of Industry and Security (BIS) announced an interim final rule targeting the remote use of advanced American AI technology. In simple terms, it's no longer enough to control where a high-end
AI chip is physically shipped. The new regulation aims to stop companies in restricted countries from simply 'renting' processing power from data centres in third-party nations. Specifically, it requires a U.S. license for any data centre in the world that provides significant computing power, derived from U.S. advanced chips, to companies headquartered in or ultimately owned by entities in places like China, Russia, and Iran.
Why Target Remote Access?
The core concern for Washington is that its existing export controls, focused on hardware sales, were being bypassed. Adversaries could gain access to powerful U.S. AI technology without ever importing a single chip. They could simply use cloud services or data centres in other countries to train their AI models. The Commerce Department stated the rule is designed to ensure sensitive American technology cannot be circumvented through these remote computing arrangements. This policy closes a significant loophole, treating remote access to computational power as a strategic asset equivalent to the physical hardware itself.
'The Know Your Customer' Mandate
A central component of this new rule is a much stricter 'know-your-customer' (KYC) or due diligence requirement for data centre operators. Essentially, data centres and cloud providers are now on the front lines of enforcement. They are required to identify their end users and understand their ownership structures. This means investigating whether a customer is directly or indirectly owned or controlled by individuals or companies in a restricted country. If a foreign data centre is unable or unwilling to provide enough information to verify compliance, exporters are expected to presume a violation is occurring. This shifts a significant compliance burden onto the global tech infrastructure industry.
Which Countries Are Specifically Mentioned?
The rule has two main prongs. The first applies globally to any data centre serving restricted entities, primarily those in countries under U.S. arms embargoes like China, Russia, and Iran (Country Group D:5). The second part of the rule introduces new, specific licensing requirements for exporting advanced chips to five countries identified as having a 'higher-than-normal risk of diversion': Singapore, Malaysia, Indonesia, Thailand, and Brazil. Shipments of advanced semiconductors to these nations will now face additional scrutiny, although exemptions may apply if the end customer is a U.S.-headquartered company. Singapore's government has stated it is studying the details and engaging with both affected companies and U.S. authorities.
Implications for Global and Indian Tech
The ripple effects are substantial for the global tech industry, including India's burgeoning data centre market. Any operator using advanced U.S. AI chips will need to implement robust customer verification processes. Customers of non-U.S. data centers will likely face new questions and contractual terms from their providers, especially if they are involved in training AI models. For data centres in hubs like Singapore and potentially elsewhere, it could create a two-tiered system, with U.S.-based hyperscalers having a potential advantage over local operators facing stricter licensing hurdles. The rule, which gives existing operators 90 days to secure necessary licenses, marks a fundamental shift in how access to AI compute is governed, moving from a focus on physical location to one based on user identity and end use.
















