Here’s what’s on our plate today:

• 🧵 Nvidia’s chips become the new geopolitical chokepoint.

• 🧠 Razzlekhan fallout, Grok deepfakes, Duolingo lock-screen shenanigans.

• 🧩 Prompt: Design an AI strategy for a mid-sized nation.

• 🗳️ Poll: Who really controls the AI chip stack?

Let’s dive in. No floaties needed…

The Laboratory

How Nvidia’s AI chips became the new geopolitical battleground

In August 1945, the world woke up to a new reality. Independent nations now held the power to destroy entire cities, wiping out millions in minutes thanks to nuclear weapons.

In the coming decades, a new force was unleashed, a race now known as the nuclear race. During this time, countries around the world developed atomic weapons, not to launch an assault on their adversaries, but to ensure their own security through the doctrine of nuclear deterrence.

Today, nine countries around the world possess nuclear weapons; however, the threat that others may be pursuing them is ever-present. It even leads to conflicts that result in the loss of lives from time to time.

In 2025, nations around the world will have another power to contend with. Apart from nuclear weapons, nations are now racing to develop artificial intelligence models that can change the very fabric of human societies.

Why AI infrastructure has become a strategic resource

In this race, companies like Nvidia that provide the base infrastructure used to run models are akin to radioactive metals like Uranium and Pultonium, both of which are essential in developing nuclear power for military and civilian uses.

The comparison is not about destructive power in a literal sense, but about strategic control: like nuclear materials, advanced AI infrastructure is scarce, dual-use, and capable of reshaping global power depending on who controls it.

Recently, in a bid to ensure that its most powerful AI chips are not used for nefarious purposes, Nvidia introduced built-in location verification technology. This, it says, can be used to indicate the country its chips are operating.

The feature, which Nvidia has quietly shown to partners in recent months, is not yet publicly available.

Behind the move are restrictions placed on the export of high-end AI chips in an effort to slow Beijing’s access to the computing muscle behind state-of-the-art AI systems.

Export controls and the limits of enforcement

The U.S. first introduced export restrictions in October 2022 and has subsequently tightened regulations to make sure the ban works. However, keeping chips out of China has proven far harder than writing the rules.

The Bureau of Industry and Security (BIS), the agency responsible for enforcing these controls, has been stretched thin. Its budget hasn’t kept up with its rapidly expanding mandate, and by some estimates, the profits from only three known smuggling operations in 2024 were larger than the agency’s entire annual enforcement budget.

Smuggling, shell companies, and legal loopholes

Meanwhile, the flow of illicit chips has grown impossible to ignore. Court filings show that from October 2024 to May 2025, a smuggling ring led by an individual named Hsu attempted to move at least 160 million dollars’ worth of restricted Nvidia H100 and H200 GPUs.

Workers reportedly peeled off Nvidia labels and replaced them with tags from a fake company called “SANDKYAN,” preparing the chips for shipment to China and Hong Kong.

In a separate case, the Indonesian telecom operator Indosat Ooredoo Hutchison bought 32 Nvidia GB200 server racks from Aivres, which contained 2,304 GPUs.

These chips ended up with a Chinese AI startup through a legal cloud service arrangement. The case showed that even legitimate business channels can still allow restricted technology to reach unintended destinations.

In response to these growing concerns, a United States senator introduced a bill that would require the Commerce Department to mandate location verification features for export-controlled AI chips.

The proposal, called the Chip Security Act, would require any regulated AI chip or any product built with those chips to include technology that can verify where it is operating. The goal is to help authorities detect diversion, smuggling, or any other unauthorized use of sensitive semiconductor equipment.

Inside Nvidia’s location verification technology

Nvidia’s new approach builds on the confidential computing features already built into its newer GPUs. These tools were created so that customers running large data centers could monitor how their chips were performing.

Nvidia has now adapted the same system to estimate where a chip is located by measuring the delay in communication between the GPU and Nvidia’s own servers. The result is a location signal that works about as well as traditional internet-based geolocation.

According to Nvidia, the agent monitors GPU health, integrity, and inventory by using telemetry and confidential computing capabilities. It then calculates a location estimate based on how long it takes for the GPUs to communicate with Nvidia’s servers.

The company stresses that the tool is read-only. Nvidia cannot send commands back to the chips, alter data, or remotely disable hardware.

Nvidia plans to introduce the location feature first on its newest Blackwell chips, which contain stronger security and attestation features than earlier Hopper and Ampere models.

The company is also looking into ways to make the tool available for older hardware.

For Nvidia, this new location technology serves several goals at once. It helps answer growing pressure from lawmakers who want proof that export controls can actually work, and it may reduce the likelihood of stricter federal mandates later.

At the same time, the feature offers practical benefits for data center operators who already need better tools to understand the health and performance of their GPU fleets across different sites.

But the move also puts Nvidia squarely in the middle of the ongoing U.S. and China technology standoff.

Geopolitical Dynamics

China reacted strongly and wasted no time making its position clear. During a recent meeting, Chinese regulators demanded that Nvidia explain what they described as possible backdoor security risks in its H20 chips that will be sold in China.

State media amplified the concerns by offering vivid examples of what could happen if foreign-made chips had hidden vulnerabilities.

One commentary in the People’s Daily described everything from an electric vehicle losing power on the highway to a remote surgery feed going dark or a mobile phone payment system failing at checkout.

The message was that a single vulnerability could trigger what it called a nationwide nightmare.

Nvidia publicly rejected the accusations. In a blog post, the company’s Chief Security Officer, David Reber, wrote that Nvidia GPUs do not contain kill switches or backdoors and that such features should not exist in the first place.

In Washington, the Chip Security Act signals growing agreement that location verification should become standard for export-controlled AI chips. The bill would require the Department of Commerce to create a system that can confirm where regulated chips are operating and detect signs of tampering or diversion. Companies would also need to report suspicious activity to the Bureau of Industry and Security.

A policy report released earlier this year, coauthored by former Google CEO Eric Schmidt, backed a similar idea. It recommended using delay-based location checks through distributed servers as a practical way to strengthen enforcement without adding intrusive hardware tracking.

Regardless of how the technology is used, it is bound to cause friction between nations that control the AI infrastructure and the ones that are working to build their independent models.

A new Cold War, this time in Silicon Valley

Today, the world stands on the cusp of a major overhaul in how technology is used; whether it will benefit all of humanity or a handful of nations will depend on how the resource, the AI chips, and infrastructure are developed and distributed.

If Nvidia’s chip tracking technology is used to ensure the technology does not fall into the wrong hands, it can help secure the future of humanity.

Otherwise, we may be entering another Cold War, one defined by chips instead of uranium, and influence instead of ideology

Bite-Sized Brains

• Bitfinex hacker Ilya Lichtenstein walks free early under a Trump-era sentencing law, reigniting questions over crypto crime accountability.

• French and Malaysian authorities are probing xAI’s Grok after reports that it helped generate non-consensual sexualized deepfakes of women.

• Duolingo tested lock-screen-style iPhone ads that appear to violate Apple’s ban on turning notifications into ad real estate.

Prompt Of The Day

Tuesday Poll

Rate This Edition