A wave of technical outages over the past week disrupted Outlook, Apple, a half dozen US banks, United Airlines and even the NSA’s classified network, and while there is no evidence the incidents were cyberattacks, security researchers say mounting AI pressure on digital defense is increasingly contributing to real-world failures. Taken individually, the incidents all appear to have routine explanations, and analysts have rejected speculation that a coordinated AI hacking spree was to blame. Microsoft said its massive 365 outage last week stemmed from a flawed update to its infrastructure that was worsened by other network failures, and officials said the NSA outage to its internal communication system came from botched patching following October’s critical PHP updates. Apple attributed its problems this week to an internal update.
The underlying dynamic researchers describe is speed. AI-assisted vulnerability discovery has produced a flood of software bugs, and attackers are using AI tools to exploit those flaws faster, pressing often under-resourced IT teams to evaluate a large number of patches and push them out to systems that countless people and organizations depend on. US senator Mark Warner said in a statement Tuesday that AI vulnerability discovery and the resulting patching are placing tremendous strain on systems that were not built to handle that pace of change. Microsoft CEO Satya Nadella said this week that more instability could be coming, because a patch needs to move instantly from availability to production once it is being exploited in the wild.
Joe Slowik, director of threat research and cyber engineering at Dataminr, said that even a couple of years ago a vendor patch for a nasty bug left some time for validation and application, but that this may no longer be an option given turnaround times that defeat patch verification entirely. Lea Schönherr, a tenure-track faculty member at CISPA Helmholtz Center for Information Security in Germany, said AI can help but people still need to understand what went wrong, confirm that a patch addresses the cause and look for similar issues elsewhere, and that rolling out patches across widely used systems takes time. Researchers also point to significant tech debt at many organizations worldwide after years of skipped investment in upgrading and maintaining digital systems.
The volume of vulnerabilities has climbed sharply. This week the total found in 2026 passed 100,000, compared with around 25,000 in 2022, the final year before LLMs went mainstream, and this year’s figure with more than a month remaining is double the combined totals of 2024 and 2025. Researchers recently disclosed AI-assisted discovery of critical vulnerabilities in PHP for the second time, in open source web server software NGINX and in the programming language Java, and Forbes reported on 35 vulnerabilities across iPhones, Macs, iPads, Apple Watches and other Apple products found by researchers using AI.
Slowik cautioned that not all AI-discovered vulnerabilities are ultimately impactful, saying the number of flaws that need to be cared about has remained somewhat constant even as the pace at which they are weaponized has increased dramatically. Only 300 vulnerabilities have been added this year to the US Cybersecurity and Infrastructure Security Agency’s catalog of exploited vulnerabilities, while VulnCheck, which runs its own database, puts known exploited vulnerabilities higher at 1,500 so far in 2026, still just 1.5 percent of the total discovered this year. Matthew Prince, CEO and cofounder of Cloudflare, predicted it will feel like the sky is falling for two years, but said AI acceleration will eventually help defenders get on top of common vulnerabilities. Schönherr said responsibility lies with the people and companies that choose to use AI and give it access to infrastructure, and that consequences and regulations should apply.
More cybersecurity news from TechManNews.







