Critical Systems Breakdowns Increasingly Frequent
A growing number of critical systems have been affected by breakdowns in recent months, impacting everything from email services to air traffic control.

Breakdowns in critical systems are becoming increasingly frequent, affecting everything from email services to air traffic control and even crypto transactions.
Microsoft Outlook was among those affected last month, as were bank servers and websites relying on PHP code.
Grocery store freezers connected to the internet have also been vulnerable, highlighting the widespread nature of these issues.
Cybersecurity experts warn that a massive cyberattack campaign or rogue AI agents are unlikely to be behind these incidents.
Instead, it appears that many of these problems are unconnected, but they all share one thing in common: an increasing dependence on fragile infrastructure.
This precarious state is reflected in the words of Adam Meyers, senior vice president at CrowdStrike, who has dubbed this month "Blacker November".
Meyers' pessimistic outlook is echoed by his assessment that things are not getting better anytime soon.
The rise of artificial intelligence has significantly altered the hacking landscape, making it easier for individuals to access and exploit vulnerable systems.
The ease with which vulnerabilities can be discovered has been significantly increased by the advent of artificial intelligence. These sophisticated models can thoroughly examine websites or code, identifying potential issues that might have gone unnoticed by human auditors. This thorough examination can be accomplished at a lower cost, allowing for more frequent assessments of even less critical systems.
As a result, numerous vulnerabilities are being exposed, earning this phenomenon an ominous moniker: the vulnpocalypse. The sheer volume of newly discovered weaknesses has created a pressing need for prompt attention and remediation from system administrators and developers.
However, the defensive advantages brought about by AI also have a darker side. Prior to the widespread adoption of these models, it could take weeks or even months for an exploited vulnerability to be identified and mitigated. Nowadays, this process can occur in a matter of days or hours, placing immense pressure on system administrators and developers.
The speed at which vulnerabilities are being discovered is leaving many organizations struggling to keep pace. Those responsible for maintaining websites, networks, and other connected systems may find themselves woefully unprepared to respond to an emerging threat. The consequences of failure can be severe, with potential liability arising from data breaches that expose sensitive customer information.
A significant factor contributing to this situation is the increasing reliance on AI-generated code. While these models are capable of producing functional software, they often lack the critical scrutiny afforded by human engineers. Companies are now employing smaller teams of human developers to oversee the work of AI agents, which can generate significant amounts of code in a short period.
The use of AI to write code has introduced new risks, as the output may not be thoroughly tested or reviewed for potential vulnerabilities. This raises concerns about the long-term reliability and security of systems reliant on AI-generated software.
John Strand, owner and CEO of Black Hills Information Security, expressed his concern during our conversation. A seasoned expert in offensive security, he noted that any instance where an attacker appears enthusiastic is typically a cause for alarm.
Strand's company specializes in testing systems for weaknesses, simulating the tactics employed by malicious actors to identify potential vulnerabilities. He reported feeling somewhat bored with the work prior to 2025, when AI models began to exhibit a significant surge in capabilities.
The rapid evolution of AI technology has led to a significant surge in capabilities, especially in the realm of hacking. According to Meyers, attackers can now use tools like Artex, which was recently highlighted by the FBI as a notable threat.
Artex is a harness that provides direction and tools for AI models to perform specific tasks more effectively. It's essentially an extension of existing AI models, allowing hackers to tap into their capabilities with greater ease.
The recent hack of major South Korean banks in September and October serves as a prime example of this growing trend. CrowdStrike determined that the attacker was likely a 26-year-old in China using Artex.
Meyers emphasizes that anyone can now become a nation-state-level hacker simply by utilizing tools like Artex. This is particularly concerning, given the vast potential for damage and disruption that such attacks can cause.
While major AI developers have implemented measures to limit access to their most capable models, hackers have found ways to exploit existing technology. Open-weight models, which can be run on individual hardware, are a prime example of this trend.
The genie is out of the bottle, as Gary Barlet from Illumio so aptly puts it. Once AI-powered hacking tools fall into the wrong hands, it's extremely difficult to regain control.
In this cat-and-mouse game between attackers and defenders, AI-powered hackers have a distinct advantage. They only need to be right once, whereas defenders must be correct every time.
Fortunately, defenders are fighting back using the same tools as their adversaries. Some programs, such as Anthropic's Project Glasswing and OpenAI's Daybreak, grant cybersecurity experts early access to capable models with fewer restrictions on what they can do.
These initiatives aim to provide a head start in patching and fixing vulnerabilities before attacks occur. By leveraging AI technology to stay ahead of the curve, defenders hope to mitigate the risks posed by increasingly sophisticated hacking tools.
The pressure to stay ahead of hackers continues to mount for tech companies and governments alike.
Big Tech leaders are well aware that current measures may not be enough to keep pace with increasingly sophisticated hacking tools.
In a recent appearance at Stanford Graduate School of Business, OpenAI CEO Sam Altman expressed doubts about the effectiveness of simply patching bugs.
According to Altman, the problem is one that has yet to receive serious attention from organizational leaders, countries, or individuals.
The situation on the ground is becoming increasingly dire, with hackers able to exploit vulnerabilities in production systems almost instantly.
Microsoft CEO Satya Nadella echoed these sentiments at the New York Times Dealbook Summit this week, highlighting the need for rapid patch deployment.
Nadella pointed out that previously, critical system patches could be tested for a week before being pushed into production. However, today's environment demands an immediate response.
The PHP vulnerability disclosed in October is just one example of the many high-profile security issues surfacing recently.
This vulnerability was made possible through work on Patch the Planet, part of OpenAI's Daybreak initiative.
The sheer volume of disclosures is staggering, with over 10,000 instances of common vulnerabilities and exposures reported to the National Vulnerability Database in November alone.
For comparison, October 2025 saw fewer than 5,000 such reports. The trend is unmistakable: a sharp increase in security issues being uncovered thanks to AI-powered analysis.
According to Matt Wyckhouse, CEO of Finite State, this surge is largely due to the ability of AI models to thoroughly examine open-source code and discover previously unknown vulnerabilities.
Wyckhouse's company focuses on connected devices, including smart home gadgets and refrigerators that may not have needed internet connectivity in the first place.
A crucial aspect of modern society's infrastructure is being threatened by a growing number of vulnerabilities in software and connected devices. These weaknesses can have far-reaching consequences if left unaddressed. The sheer volume of issues that need to be patched is overwhelming companies that specialize in software creation or operation.
Frequent updates are becoming a norm for computer systems, as they strive to stay ahead of potential threats. However, this constant cycle of patching and updating is putting immense pressure on the teams responsible for maintaining these systems. The risk of an attacker exploiting a vulnerability before it can be patched is ever-present.
Ted Miracco, CEO of Approov, has expressed concern about the sheer scale of work required to rectify the situation. He believes that tackling this backlog will take considerable time and effort. This sentiment reflects the frustration felt by many in the industry who are struggling to keep pace with the evolving threat landscape.
Not all patches are effective, and sometimes they can even cause problems. In some cases, software updates have been identified as contributing factors to recent high-profile outages, including Apple's issues last Tuesday. This highlights the complexity of the issue at hand.
The intersection of AI-generated code and existing vulnerabilities is a growing concern. When models trained on vast amounts of data are used to write new code, they can produce functional but flawed software. This raises questions about the long-term implications of relying on AI for coding tasks.
Software engineering has become one of the industries most disrupted by AI's increasing presence. As companies seek to reduce costs and boost their stock prices, some have turned to using AI-generated code instead of employing human engineers. However, this shift may ultimately lead to more problems than it solves.
The recent trend of Big Tech companies laying off staff on a large scale is also worth noting. Microsoft's decision to offer buyouts to 60,000 employees has sparked concerns about the implications for system maintenance and reliability. Some experts believe that having more people on hand to maintain these systems might actually lead to fewer breakdowns in the long run.
The increasing reliance on AI in software development is leading to new challenges, as junior developers are now taking on more responsibilities without the guidance of senior colleagues.
As a result, vulnerabilities are being written into new code and patches, which can have far-reaching consequences for companies and individuals alike.
This situation highlights the complex relationship between AI and software security - while AI can provide solutions, it also creates problems that must be addressed in order to achieve greater stability.
In the short term, it seems that patience will be a necessary virtue, as companies navigate this period of growing pains and wait for the benefits of improved software security to materialize.
Facts based on reporting originally published by CNET.
You may republish this story, in full or in part, if you credit News Central Site and link to it (licence CC BY 4.0). Photos are not included.



