Centralized vs Decentralized Internet: When AI Automation Goes Wrong
When big parts of the internet go down, the world seems to pause. That’s what happened during many incidents over the last year, when a tiny software glitch, update, or DNS issue caused a global chain reaction. Within minutes, food deliveries stopped, airports dived into chaos, and hospitals lost access to critical systems. These types of events showed just how vulnerable the modern web has become to single point of failures.
Today — The centralization
Today, most online traffic runs through a few tech giants like Amazon, Google, Microsoft, cached with Cloudfront and Cloudflare. It is easy, scalable and best in class, but it also concentrates all our hopes in terms of reliability in very few systems. When one of those systems breaks, it turns into a real snowball disaster and a digital earthquake across the planet.
Tomorrow — Decentralization is required
In a decentralized web, no one holds all the keys. The system control is shared across all the independent networks. Imagine that every home generates its own solar power instead of relying on one massive power plant. If one house would go dark for some reason, that would not impact the others. Technologies like blockchain, IPFS, and federated platforms point toward that kind of future, where the internet could keep running even if parts of it went down. That’s something we might need to embrace as soon as possible.
The Hidden Risk in AI Automation
Automation quietly runs much of the online world: routing data, load-balancing servers, applying updates, and keeping traffic smooth. But when automation fails, it fails instantly. A small configuration error can take down thousands of services at once. In centralized setups, shared dependencies make those errors spread fast. In decentralized systems, the problem would stay contained and isolated.
AI Automation and the Human Factor
Sadly, we often see that the giants impacted by those worldwide outages are also often the ones laying off massively. The move always seems promising and efficient at first. Replacing the human in the loop. But when the network crashes, the lack of experienced engineers becomes clear. Senior staff members often carry institutional knowledge, the kind of instinct that can’t be programmed. Those subject matter experts will be the one to notice small irregularities that signal bigger failures. Without them, recovery takes hours instead of minutes, and you might argue that in the future AI agents’ orchestration will solve this. Will it really make a difference?
When AI Makes the Web More Brittle
Ironically, the AI tools built to prevent such failures make things worse. Automated recovery systems try to restart systems sometimes too aggressively, overloading servers before engineers could slow them down. In recent issues, analysts described it as “AI brittleness,” where intelligent systems misfire because of a lack of context.
The Takeaway
Those outages are more than a technical hiccup. It is a warning about concentrating too much power and too much automation in too few hands. The next time an app goes dark, remember that the issue often lies not in the tools themselves but in how tightly they are controlled and interconnected. A more resilient internet means spreading both technology and responsibility instead of keeping them confined to one corner of the cloud in the hands of AI agents instead of humans in the loop.
Read the full article here: https://dimitripletschette.medium.com/centralized-vs-decentralized-internet-when-ai-automation-goes-wrong-939b27c77dba