Artificial intelligence
Recent outages were not linked to a cyberattack, but AI is making vulnerabilities faster to find and exploit Igor Omilaev/Unsplash

A series of technology disruptions affecting major companies and essential services has highlighted the vulnerabilities of increasingly interconnected digital systems, as artificial intelligence changes the way security weaknesses are discovered and exploited.

Authorities and affected companies have reported no evidence that the recent outages were connected to a coordinated cyberattack.

However, cybersecurity experts warn that increasingly sophisticated AI tools are changing how quickly vulnerabilities can be identified and exploited. The resulting rush to secure affected systems can also create operational challenges, particularly when organisations temporarily suspend services to install emergency updates.

A String of Outages Hits Major Global Services

The disruptions reported during the week included problems at Microsoft, JPMorgan Chase, Apple, United Airlines, Marriott and the US National Security Agency (NSA). The incidents affected everything from workplace communications and online banking to flight operations and access to digital services.

Microsoft 365 users experienced interruptions to Outlook, Teams and Copilot after a network change exposed a previously existing defect in a fix the company had begun deploying over the summer.

Separately, a malfunction in the Federal Aviation Administration's network triggered a nationwide ground stop lasting more than four hours. United Airlines experienced another disruption several days later, forcing it to delay departing flights.

JPMorgan Chase reported transaction delays linked to a software bug, and Apple attributed disruptions to an internal update. Authorities and the companies involved said they had found no evidence connecting the incidents to a cyberattack, according to Axios.

AI-Detected Security Flaws Trigger Emergency Fixes

One example of how cybersecurity improvements can disrupt online services emerged on 21 October, when organisations around the world rushed to address serious vulnerabilities in PHP, a programming language used to build websites. AI systems operated by researchers helped identify the flaws, which were considered capable of exposing affected websites to data theft and disruption.

The vulnerabilities, dubbed PHP2Shell, prompted emergency security measures across several industries. Banking, airline, telecommunications and government services were among those affected, alongside websites including Wikipedia and parts of the BBC, according to the World Economic Forum.

The US Cybersecurity and Infrastructure Security Agency urged organisations to install the necessary fixes or take vulnerable systems offline until they could be secured. The incident illustrated how identifying a security weakness can create an immediate operational challenge for organisations responsible for keeping essential services running.

Hackers Are Also Learning To Exploit AI

The concern is not limited to accidental disruptions caused by security updates. Criminal groups are also exploring how artificial intelligence can help them identify weaknesses in computer systems and prepare attacks more quickly.

Google previously reported disrupting an attempted cyberattack in which criminals used an AI model to help uncover a previously unknown vulnerability in a widely used system administration tool. The flaw could have allowed attackers to bypass two-factor authentication.

Google said its Threat Intelligence Group worked with the affected software vendor to disclose the vulnerability responsibly and disrupt the suspected criminal activity before it could be used in a planned mass exploitation campaign.

The incident showed why cybersecurity specialists are increasingly concerned about the speed at which AI can assist malicious activity. Google's threat intelligence team warned that criminals were moving beyond basic chatbot prompts and using more advanced techniques to identify and exploit weaknesses. These capabilities could help attackers pursue data theft, ransomware and extortion before organisations have enough time to respond.