Outages have hit Outlook, Apple, six US banks, United Airlines, and the NSA’s classified network. While no evidence points to a coordinated cyberattack, speculation is running wild. Some believe artificial intelligence is the culprit. That is not the case. Each incident has a routine explanation. However, security researchers now agree that AI is starting to cause real-world failures.
In this article
The speed of change
AI tools find software bugs faster than ever. Attackers use the same tools to exploit them quickly. This pressure forces organisations to patch systems rapidly. IT teams often lack the resources to evaluate every update or push it to systems used by countless people. The core impact is simple: AI speeds up everything.
Microsoft’s 365 outage last week stemmed from a flawed infrastructure update worsened by network failures. Officials stated the NSA outage resulted from botched patching following critical PHP updates in October. Apple attributed its issues to an internal update. US senator Mark Warner said on Tuesday that AI vulnerability discovery and the resulting patching are putting tremendous strain on systems not built for that pace of change. Microsoft CEO Satya Nadella added this week that a patch must move instantly from availability to production because it is being exploited in the wild.
External pressures and bugs are not new. Many organisations carry significant tech debt. They have cut corners and failed to upgrade digital systems for years. This practice was always risky. Companies without resources or the will to invest in security often gamble with outdated systems and missing safeguards, hoping to avoid the worst consequences.
Joe Slowik, director of threat research and cyber engineering at Dataminr, says: “Historically, even a couple of years ago, a vendor releases a patch for a nasty bug. You might have some time to do some validation and apply it.” He notes that turnaround times now defeat the ability to verify patches.
IT administrators rely on AI to accelerate evaluation and prepare systems for updates. Current tools cannot replace human analysis and planning. Lea Schönherr, a tenure-track faculty member at CISPA Helmholtz Center for Information Security, a German national research institution, says: “AI can help, but people still need to understand what went wrong, check that the patch addresses the cause, and look for similar issues elsewhere.” She adds that rolling out patches across widely used systems also takes time.
Matthew Prince, CEO and cofounder of Cloudflare, predicts the situation will feel like the sky is falling for two years. He suspects AI acceleration will eventually help defenders manage common vulnerabilities and secure software better. “These new threats and new security vulnerabilities won’t happen as often, and the background noise will actually tick down quite a bit,” he predicts.
Numbers are rising
The evidence that AI has fundamentally changed cybersecurity is mounting. This week, the total number of vulnerabilities found in 2026 surpassed 100,000. In 2022, the final year before large language models went mainstream, there were around 25,000 recorded vulnerabilities. This year’s figure, with more than a month still to go, is double the combined totals of 2024 and 2025.
Researchers recently disclosed using AI to find critical vulnerabilities in PHP, open source web server software NGINX, and the programming language Java. Forbes also reported on 35 vulnerabilities across iPhones, Macs, iPads, Apple Watches, and other products from the company discovered by researchers using AI.
Not all AI-discovered vulnerabilities are impactful. Slowik says: “The number of vulnerabilities has certainly exploded, but the number of vulnerabilities that you need to care about has remained somewhat constant.” He adds that the caveat is the pace at which they are used or weaponised has increased dramatically.
Only 300 vulnerabilities have been added to the US Cybersecurity and Infrastructure Security Agency’s catalog of exploited vulnerabilities this year. This means the flaws hackers successfully used are far lower than the total discovered. VulnCheck puts the number of known exploited vulnerabilities higher, at 1,500 so far in 2026. This is still just 1.5 percent of the total number of bugs discovered this year.
What matters is avoiding blame when things go wrong. Schönherr says: “Responsibility lies with the people and companies that choose to use AI and give it access to infrastructure, and there should be consequences and regulations that can be applied.”
Other news
• JPMorgan Chase confirmed hackers stole some company data in a separate incident. The bank says attackers accessed certain marketing data held by third parties. This may include customer information such as names and emails. The security incidents are not linked to the earlier outages. Visa and Mastercard also had data impacted in attacks against their vendors.
• Police in the Netherlands revealed they thwarted several attempts at violence directed toward customers of the digital bank Revolut. In September the fintech company was reportedly compromised, with customer data stolen. The National Police Corps published a statement saying they had made several arrests which can be linked to the data theft. It is unclear who was being targeted, but there has been an increase in physical attacks against wealthy cryptocurrency holders over the last few years.
• The UK government has been forced by a court ruling to admit it ordered Apple to circumvent its own encryption and access customer iCloud data. British officials had maintained it would damage national security to confirm they issued Apple a technical capability notice last year. Under British laws, the notices are supposed to be shrouded in secrecy and not talked about in public. However, the existence of the order was widely known, with top Trump officials publicly criticising it and UK government officials allegedly briefing newspapers about it.
What it means
Organisations cannot patch fast enough to keep up with AI-driven discovery. The volume of bugs is overwhelming, even if most remain unused. Human oversight is still required to verify fixes and understand root causes. Companies must address tech debt or face constant instability.




