AI found a massive PHP flaw. The rushed fix broke the NSA and US flights.

Major US outages weren't a cyberattack. AI found vulnerabilities so fast that rushing to patch them crashed the systems they protect.

Drama ยท Source: Gizmodo

What happened

America's digital infrastructure choked this week. Microsoft Outlook and Teams went down for nearly 24 hours. The FAA ordered a nationwide ground stop that delayed thousands of flights for over four hours. JPMorgan Chase customers could not access their accounts or transfer money. People immediately panicked about a coordinated cyberattack. Homeland Security Secretary Markwayne Mullin had even warned about Iranian hackers targeting critical infrastructure.

The truth is actually weirder. Director of National Intelligence Jay Clayton confirmed the outages stem from a frantic rush to patch software. AI models recently discovered massive vulnerabilities in PHP. That programming language powers roughly 70 percent of the world's websites. The flaws could allow hackers to read memory containing sensitive passwords. Researchers used AI tools to exploit these PHP flaws. This triggered a global panic to fix them.

Wall Street executives and the NSA told Congress that companies pushed too many patches at once. The systems simply broke under the weight of the updates. The NSA itself suffered an outage on its highly classified internal communications network called NSANet. The agency had to issue a code yellow and then a code orange. They closed the Fort Meade campus to most employees because a routine software patch failed. The rush to patch AI-discovered vulnerabilities is breaking the very systems they are supposed to protect.

Key facts

Why it matters

AI operates at machine speed. Human IT infrastructure does not. We now have frontier models that can find and exploit zero-day vulnerabilities faster than legacy systems can handle the updates. Cybersecurity expert Nicole Perlroth noted we depend on software almost nobody understands end to end. Attackers can now probe systems at machine speed. Defenders are trying to respond just as fast. But enterprise systems are fragile. When you force a massive volume of patches through aging telecommunications architecture, the cure becomes worse than the disease. The FAA traced its massive failure to a single router in Arkansas overwhelmed by network traffic.

This changes the core cadence of cybersecurity for anyone building software. Finding the flaw is no longer the bottleneck. The real bottleneck is safely deploying the fix. Senator Mark Warner pointed out this is just the consequence of how quickly AI is changing the landscape. If companies keep panic-patching every AI-discovered vulnerability at once, self-inflicted outages will become the new normal. You have to balance the risk of getting hacked against the risk of taking your own servers offline.

For builders

Stagger your security patches

Do not push massive updates all at once. The NSA and major banks crashed their own networks by rushing to fix AI-discovered flaws. Build staggered rollout pipelines to catch breaking changes early. Your infrastructure team pays the price if you rush.

Audit your PHP dependencies

The root cause of this week's chaos was a vulnerability in PHP. If your stack relies on legacy PHP, you are in the blast radius. Allocate engineering time to update vulnerable components safely. Founders lose revenue when rushed patches take down production.

Build redundancy for outages

The NSA survived its network crash because it had multiple backup systems to communicate with the President. You need fallback plans for when your primary tools fail. Customers will churn if a botched patch locks them out of your product.

My take

We spent years worrying AI would hack us. Instead, AI found our flaws so fast we hacked ourselves trying to fix them. If your deployment pipeline cannot handle rapid patching without crashing, you are already behind. Stop panic-pushing code and start building resilient systems.

Original reporting: Gizmodo. This is my rewrite and opinion.

More AI news for builders