A few hours ago I tried to rent a Lime bike to reach a therapy appointment. I opened the app, scanned the QR code, and authorized Apple Pay. The app hung briefly and then closed. I tried several more times: same result. Uber could not take payment either. DownDetector made the picture clear: Apple Pay, which I thought lived only inside the phone, is a cloud service and it was down. No bike; I jogged.
That single outage is trivial. But the way the mind searches for a cause โ broken phone, dead network, cloud service, cyberattack โ is the same pattern friends and colleagues have been applying for weeks to everything that collapses: planes, banks, Microsoft. The most common answers are โweโre being attackedโ or โAI went rogue.โ The more uncomfortable explanation is that digital infrastructure is showing structural fragility, not a coordinated attack. Even a cyberattack against the International Ice Hockey Federation, which would normally pass unnoticed, becomes part of the same story.
Over the past few years, mass layoffs of developers and technical staff, deep cuts to public oversight, loss of institutional knowledge, underinvestment in redundancy and maintenance, and a rush to automate with LLMs and generated code have piled up. On top of that, LLM-assisted vulnerability discovery and exploitation force companies to ship patches quickly, often on live systems. The NSA described temporary degradation during routine maintenance; CISA says more resilience is needed. But resilience requires people who understand how systems can fail.
The PHP2Shell exploit, discovered with LLM assistance, knocked sites offline as operators tried to update faster than usual. Calif announced serious Java vulnerabilities, also found with LLMs. The time between discovery and exploitation has shrunk, and the response is often a rushed patch on production infrastructure, including air traffic control.
When Elon Musk took over Twitter and renamed it X, many expected a collapse from lost expertise. It did not happen exactly as predicted, but that philosophy โ fire the experts and hope AI patches the holes โ has become an explicit strategy in parts of the economy and government. The result is not one incident but a steady stream of failures that remain isolated only until someone connects them.
The point is not whether AI is guilty. Many outages have nothing to do with AI. The point is the loss of control and understanding of complex systems. Jimmy Breck-McKye describes engineers who can no longer solve problems without AI assistance, not because they are incompetent but because of complexity: a wave of generated code produces small, incoherent changes that look rational in isolation but make no sense as a whole. It is the same dynamic as a Ship of Theseus in which every piece is replaced and nobody sees the overall structure anymore.
For those running local stacks or evaluating self-hosted deployment, this is not just a cloud problem. Apple Pay stops working because its infrastructure is distant from the user, but the core failure is not the cloud itself: it is dependence on components that no one inside the organization can read and repair. Redundancy cannot be activated with a subscription. It requires the opposite: local control, people who can diagnose, and awareness that AI accelerates both bug discovery and the production of fragile patches. AI-RADARโs /llm-onpremise analytical frameworks are one way to evaluate these trade-offs between TCO, sovereignty, and control.
There is no need to imagine a coordinated cyberattack to explain the disorder. It is enough to accept that we are replacing pieces of the economy and infrastructure without understanding how they work together. For now, we should keep expecting bad things to happen.
๐ฌ Comments (0)
๐ Log in or register to comment on articles.
No comments yet. Be the first to comment!