The Internet began to break down all at once, but the cyber apocalypse is still far away – and here’s why

Источник изображения: Zan Lazarevic / unsplash.com

Over the past week, several major technology incidents have occurred: Americans temporarily lost access to their bank accounts, passenger planes stopped flying, and even email stopped working. So far, all these incidents are just a coincidence – a large-scale cyberpocalypse has not yet arrived, experts reassure.

  Image source: Zan Lazarevic / unsplash.com

Image source: Zan Lazarevic / unsplash.com

The past week was rich in cyber incidents, data from the Downdetector platform confirms, but if you take a sample of 30 days, no statistical anomalies are observed. However, they have a direct connection to artificial intelligence, US national security officials and Wall Street said. Massive system failures were caused by attempts to correct software errors detected by AI systems. Large organizations are in an all-out race to avoid “buggeddon,” a scenario in which AI-powered hackers find and exploit vulnerabilities in the Internet before they can be patched. This process became the cause of instability.

Last week, Microsoft was unable to resolve a massive outage in its systems for more than 24 hours; The US Federal Aviation Administration (FAA) suspended flights across the country for several hours due to overload on routers; and on Monday, November 16, some Americans lost access to JPMorgan Chase banking services. Homeland Security Secretary Markwayne Mullin suggested that AI systems that can write and modify code could have contributed to the failures. However, no evidence of this could be found: in most cases, failures are explained by software updates, due to which small technical problems grew into large-scale ones.

JPMorgan Chase attributed the problem to a software error that delayed transactions involving other banks; The FAA said the flood of network traffic destabilized legacy equipment. Microsoft talked about an error that occurred several months ago, an increase in traffic and a decrease in server capacity. Organizations around the world are looking to update software to improve system reliability, given that AI models are able to find and exploit vulnerabilities at high speed.

This race, experts point out, has sped up much of the coding work, a process that can lead to increased downtime, operational problems and unexpected failures. AI has also significantly increased the speed at which engineers write and improve software, leading to new bottlenecks in testing the safety and reliability of new code.

In the seven days from November 9 to November 16, Downdetector recorded nearly 6 million reports of problems in the US, the highest weekly figure for the year. But there were only three major outages in 30 days, compared with the usual five.

If you notice an error, select it with the mouse and press CTRL+ENTER.