Juan Londoño and Jennifer Huddleston
The week between November 11 and 17 was plagued by various newsworthy outages of digital and telecommunications services. From Microsoft and the Federal Aviation Administration to online banking services, Americans faced life-disrupting outages. The rapid succession led many to raise concerns and speculate about potential cyberattacks or overreliance on artificial intelligence (AI). But in many cases, the actual cause turned out to be simpler and less exciting than speculated.
There is still a need for more clarity and facts to understand what happened with each outage. At this time, they do not seem to be connected to a single cause. Some were due to routine internet malfunctions or issues from updating faulty equipment. Other issues, like those experienced by banks and the National Security Agency, were due to a rapid succession of security patches that addressed important software vulnerabilities discovered by AI.
While public skepticism may have been quick to blame AI, the opposite is true in some ways. AI may have prevented a bigger problem in the future. As cyber-capable AI becomes more ubiquitous, its defensive applications are increasingly important and need bolstering. But the quick speculation about the fallout from some of these outages highlights how the general animosity toward AI can be counterproductive to technological progress and cyber response efforts.
Why were so many quick to presume AI was to blame? Some of this animosity seems intentional, or at least coordinated, as recent reports indicate that bot activity on social media during the outages peddled conspiracy theories about AI being the culprit behind them. But at the same time, AI is being used to discover major vulnerabilities in essential internet infrastructure and the software of critical services like online banking. As AI-enabled cyberattacks rise, driven by publicly available models like ARTEX, defensive AI tools are becoming increasingly important. While it is important for companies and individuals to engage in up-to-date cybersecurity practices given the risks, focusing solely on the problem may obscure the beneficial role AI plays in responding to these concerns, as well as how these cybersecurity issues predated our current AI technology.
The biggest downside of this general rush to portray AI negatively is that it creates pressure on policymakers to move away from a more sensible approach to policymaking toward heavy-handed regulation or outright bans on development. But in the meantime, AI is finding critical vulnerabilities in some of the building blocks of the internet and making everyone safer in the process. In October 2026, AI tools discovered a vulnerability in PHP, a programming language used in about 70 percent of websites. The exploit was so grave that many websites, including various government sites, went offline to fix the issues rather than risk serious cyberattacks. Less than a month later, AI discovered a vulnerability in another major programming language, Java, which is used by most US government agencies and Fortune 500 companies. Without AI discovering these flaws, we could have been looking at even more significant outages or vulnerabilities.
Outages are inconvenient for users and can cause problems beyond the impacted site or service, as many experienced in November 2026. But the existence of offensive AI tools, especially those that are public and open-source, like ARTEX, makes AI cyberattacks ever more likely. Meanwhile, much of the current software and website design has been built on legacy systems and code that are unlikely to be vetted by defensive AI tools. Leaving these vulnerable systems online would expose civilians to a tremendous risk of attack. These outages and service interruptions are more of a byproduct of the growing pains of the widespread adoption of AI than an indictment of the technology itself.
To outmaneuver AI-enabled attackers, cyber defenders need access to tools more advanced than those used by bad actors. The only way to give defenders these tools is to ensure frontier development continues and that they have access to continuously improved models. Allowing the technological gap between attackers and defenders to narrow will only lead to more cyberattacks, which in turn will lead to data leaks, economic losses from ransomware payments, or—ironically—more outages.
The animosity and negative sentiment around AI can lead the public to rapidly blame AI when technology goes awry. Unfortunately, this may tempt policymakers to respond with calls for stricter regulation. But heavy-handed regulation can threaten the development of those defensive tools and fail to solve the underlying issue of cybersecurity vulnerabilities. Policymakers face the difficult task of cutting through the noise, ensuring cooler heads prevail, and not hampering the development of a technology that is becoming increasingly important to users’ welfare and safety. Yes, AI can cause new cybersecurity problems, but it is also an important part of our cybersecurity arsenal.
