Wednesday, July 8, 2026
Subscribe

The AI Cold War Has Entered Your Home Router

Nation-state hackers may already be inside 20 million household devices. The question is whether governments or businesses can move fast enough to stop them.

Jun 23, 2026 · 18 Minutes

A Cold War That Lives in Your Living Room

The conventional image of a geopolitical conflict involves warships or sanctions. The new version involves your smart speaker, your home router, and an audio clip you cannot hear but your AI assistant can.

This week's episode of Good Revenue covers a lot of ground fast, but one thread ties almost everything together: the US-China technology cold war has stopped being a story about chip factories and export licenses and has become something far more intimate and far harder to contain.

The 96% Problem

Start with the most unsettling data point. A research team in China presented findings at a US conference just two months ago showing they had found a way to embed hidden commands inside digital audio waveforms. These commands are inaudible to any human ear. The AI voice models that receive them, however, interpret the signal perfectly. The reported success rate: up to 96%.

This is not a theoretical vulnerability. It targets large audio language models, the same category of technology that is being embedded into customer service systems, consumer devices, and enterprise software at speed. And because the attack modifies numerical values inside the waveform rather than inserting obvious noise, it is exceptionally difficult to detect, let alone filter.

Security researchers who reviewed the findings have described it as more concerning than a traditional prompt injection attack, because there is no visible text, no obvious trigger, and no straightforward audit trail.

Five Eyes Is Worried, and That Matters

Warnings from intelligence agencies tend to be vague by design. The recent joint statement from Five Eyes, the intelligence-sharing group comprising the US, UK, Australia, Canada, and New Zealand, was unusually pointed. Without naming specific countries, the group told business leaders and government officials to treat AI-assisted, state-sponsored cyberattacks as a present and serious threat, not a future one.

Reporting from the Wall Street Journal adds texture to why. Nation-state actors have been found to be hijacking everyday consumer electronics, building proxy networks to mask their traffic, and exploiting devices in ways that remained undetected for years. In one case, a conversation between a Microsoft executive and a Comcast counterpart, initiated to chase down some suspicious IP addresses two years ago, eventually unraveled a Russian-linked operation of a scale that reportedly alarmed even veteran cybersecurity professionals. The estimate from that reporting: something in the range of 20 million devices could be implicated.

This is not a problem with an obvious fix. The compromised devices are already in homes and offices. The attack surfaces are multiplying as more hardware connects to the internet.

The Broader Escalation

The cyber dimension is just one front. The episode this week also tracks the formal, institutional escalation happening in parallel.

The US Department of Defense expanded its list of banned Chinese companies in early June to include firms that would not traditionally be classified as defense contractors, among them Alibaba, Baidu, and BYD. The reasoning is that these companies are considered sufficiently intertwined with the Chinese state that doing business with them carries national security implications. Beijing's response was swift: 10 US firms were cut off from Chinese rare earth materials and dual-use technology components, a move aimed squarely at the drone and defense manufacturing supply chain.

Rare earths are worth pausing on. China is the dominant global producer, and these minerals are critical inputs for everything from defense hardware to consumer electronics. The export restrictions China began imposing roughly a year ago were intended as leverage. The unintended consequence, which Neeta flags in the episode, is that they have dramatically accelerated investment in alternative sources in Australia and elsewhere. Pressure can create its own supply.

ASML and the Smuggling Question

Complicating the picture further is the situation around ASML, the Dutch company that manufactures the extreme ultraviolet lithography machines essential for producing the world's most advanced chips. The US Commerce Secretary met with ASML executives last week to raise concerns, reported by Bloomberg and Reuters, that one of these machines may have reached China.

ASML's denial is careful and specific: it states the company has never shipped an EUV machine or components designed specifically for one to China. What the statement does not address is the well-documented pattern of high-value technology being routed through third countries, including Malaysia, Thailand, and Singapore, to circumvent export controls. The episode does not assert that this happened here. But the question of what the statement does not say is at least as important as what it does.

What Comes Next

SpaceX's turbulent first week as a public company, down roughly 16% before a partial recovery, adds another variable. On day 15 after the IPO, the stock enters the Nasdaq indices and early retail investors gain their first opportunity to sell. How that plays out will be a useful read on whether enthusiasm for high-profile AI-adjacent listings has genuine staying power.

The bigger picture, though, is that AI has fused with geopolitics in ways that are proving very difficult to unpick. Technology decisions that once lived in procurement spreadsheets now carry foreign policy consequences. Devices people buy at retail have become potential instruments of state-level conflict. And attacks on AI systems can arrive in formats that are, quite literally, beyond human perception.

The episode closes without easy reassurance, and that feels about right.

Sources & Further Reading
Watch Next
Subscribe Free →