Two Fronts: AI Models & Critical Chips Threaten Supply Chains

The world's digital infrastructure is facing a quiet but profound shift in its underlying vulnerabilities. For years, we've worried about physical components – the chips, the raw materials, and the factories that produce them. But new concerns are emerging, not just about where our critical hardware comes from, but about the opaque, often unpredictable behavior of the artificial intelligence models we increasingly rely on.

Today, May 20, 2026, two distinct but related alarms sound in the tech world. Keeper, in a piece titled "When Models Eat the World," points to the growing instability introduced by third-party AI models. Meanwhile, Daniel Chiang reports from Seoul for DigiTimes, highlighting South Korea's near-total dependency on foreign photonic semiconductors for its defense systems. Both articles paint a picture of strategic dependencies that could have far-reaching consequences.

The Invisible Hand of AI Models

For most of software history, a supply chain issue meant a missing library, an API change, or perhaps a security flaw in an open-source component. We could, for the most part, inspect the code, understand its logic, and anticipate its behavior. AI models, however, are different.

Keeper argues that when a significant portion of your software's quality hinges on a model built and controlled by someone else, whose internal workings are a black box, you're in uncharted territory. These aren't static libraries; AI models can, and do, change their behavior without warning. This isn't always due to explicit updates, either; ongoing training or reinforcement learning can cause a model to spontaneously shift its internal logic based on external, unmonitored data. Imagine a critical component in your product pipeline suddenly deciding to act differently without you ever touching its code. It's a nightmare for reliability, security, regulatory compliance, and even legal liability.

This new layer of uncertainty complicates everything from product development to national security. How do you audit a system whose fundamental decision-making engine is constantly evolving? How do you guarantee a certain level of performance or ethical behavior when the underlying logic is a moving target? Companies are only just beginning to grapple with this fundamental lack of control over key parts of their digital infrastructure.

Hard Dependencies in a Digital Age

While AI models present a novel, software-based supply chain risk, the old anxieties about hardware remain very real. Daniel Chiang's report from Seoul brings this into sharp focus, detailing South Korea's precarious position in the global photonic semiconductor market. These chips are no longer niche; they're indispensable for modern defense, prized for their ultra-high-speed data processing, high capacity, low power consumption, and exceptional reliability. Think advanced radar, secure communications, and precision guidance systems.

Despite their strategic importance, South Korea is staggeringly import-dependent, sourcing 99% of its photonic semiconductors from abroad. This isn't just an economic issue; it's a critical national security vulnerability. In a global crisis, or even a targeted geopolitical dispute, access to these essential components could be severely curtailed, crippling the nation's ability to defend itself.

We've seen how fragile global chip supply chains can be. The automotive industry, for instance, took a significant hit during the COVID-19 pandemic due to chip shortages. For military applications, the stakes are exponentially higher. This reliance isn't unique to South Korea, but it serves as a stark reminder of how deeply intertwined national security is with specific, often geographically concentrated, technological capabilities.

Why it matters

Taken together, these two reports paint a worrying picture of our increasingly complex technological dependencies. On one hand, we're losing control over the very algorithms that drive our software. On the other, nations remain dangerously beholden to foreign sources for the hardware that underpins their most critical systems. The traditional notion of a supply chain is expanding, now encompassing not just physical goods, but also the invisible, ever-shifting logic of artificial intelligence. Businesses will need to push for greater transparency and auditability from AI model providers, while governments must urgently reassess their strategic dependencies, investing in domestic capabilities for critical technologies. The future of technological autonomy, and indeed national security, hangs in the balance.