AI-driven vulnerability discoveries and rushed software updates fueling outages

According to CyberScoop, federal officials and lawmakers are pointing to accelerated software patching as a possible contributor to recent technology outages affecting airlines, banking systems and major communications platforms.

At a November 18 congressional hearing, intelligence and homeland security officials said there was no evidence that the disruptions resulted from malicious cyberattacks. Instead, they cited the increasing pace of software updates and system upgrades.

Sen. Mark Warner warned that AI is enabling organizations to discover vulnerabilities at unprecedented speed, creating pressure to deploy patches immediately, sometimes across systems not designed to handle such rapid changes.

The concern is that emergency security updates, particularly those affecting critical infrastructure, may introduce operational failures when deployed without sufficient testing.

Ryan McCurdy, Vice President of Marketing, Liquibase

“AI is changing how quickly organizations can discover vulnerabilities but fixing them safely is still a complicated process. Engineering teams need to understand dependencies, test changes, and deploy fixes into systems that millions of people rely on. That work doesn’t become less important because the threat is urgent.

“I think the danger is treating speed as the only measure of success. A patch can close a security vulnerability and still create an outage if it breaks something else. We’ve seen several disruptions involving software defects, infrastructure changes, and maintenance, although we can’t attribute all of them to AI-driven patching.

“As we move toward an AI-driven software development lifecycle, enterprises need to automate the controls that make changes safe, not bypass them. For production databases, that means validating changes, enforcing policies, requiring approvals when necessary, and maintaining audit evidence of what happened. The goal shouldn’t be to patch slower. It should be to make safe changes faster.”

ㅤ

Darin Fredde, Senior Director and Offensive Security Evangelist, Ridge Security

“Accelerated patching is a plausible explanation, but it has a familiar ring to anyone who’s heard an airline announce a ‘software glitch.’ I’d give that explanation more weight once we know what changed, how the failure spread, and why the fallback didn’t hold.

“When a handful of shared providers can disrupt airlines, banks, and communications platforms, I think about the A in the CIA triad: availability. AI can help teams find flaws faster, and some flaws genuinely warrant emergency action. The discipline is knowing which fixes need to go out now, rolling them out in stages where possible, and having a tested way back. Otherwise, it starts to look a little like the dodos in Ice Age: everyone’s focused on saving the watermelon, and the cliff sneaks up on them.

“The practical question I’d ask is whether we can still deliver essential services if a shared provider goes down. Can we route around the failure? Can staff keep critical work going offline if a cloud service is unavailable? Those are things we can test today, even while the incident reviews are still underway. That’s where I’d put the next dollar. One question I’d add to any outage drill is whether you can still test your defenses with the internet connection down. That’s one reason we built RidgeOne. It runs security tests and the AI model locally, so teams can keep testing inside their own network without a live connection to a cloud service. The cloud can go down. Your security testing can keep going.”

Organizations would do well to get ahead of this by not trusting AI as they are now. Because trusting AI clearly isn’t working.

Leave a Reply