Why Tech Outages Are Piling Up, and What AI Has to Do With It
Apple Pay, the FAA, NASA, and Microsoft all had outages in two weeks. Officials say no cyberattack, and experts offer a different explanation involving AI and old systems.
A run of major tech outages hit U.S. organizations over the past two weeks, and TechCrunch reports that some security experts see AI as part of the explanation. Apple Pay, NASA's internal network, FAA flight operations, and Microsoft's Outlook, Teams, and Copilot were all affected. Officials and companies say these were not cyberattacks, but the theories about what really happened offer useful lessons for any organization that runs on software.
What went down in the past two weeks
According to TechCrunch, Apple Pay went down, and so did NASA's internal communication network. The FAA grounded all flights for hours, which TechCrunch says has happened only three times since 9/11. Microsoft had an outage that took down Outlook, Teams, and Copilot for nearly two days.
The obvious first guess is an attack. TechCrunch notes that hackers recently hit refrigerators in a grocery store and a blood bank in France, which made that guess feel natural. But the director of National Intelligence said the U.S. outages do not appear tied to state-sponsored terrorism, and President Trump called them "individual events." Many of the companies involved have said the cause was technical problems, not cyberattacks or AI attacks.
The timing is what keeps people asking questions. The outages came close together, and only a few weeks after what TechCrunch calls cybersecurity's "Black October." On October 21, a number of separate security incidents struck PHP, Node.js, LiteLLM, and others.
Why some experts link these tech outages to AI
Dave DeWalt, an investor who spent more than three decades in cybersecurity (including as CEO of McAfee and vice chairman of the CISA Advisory Council), told TechCrunch he thinks the outages are probably AI-related. His view is that they are the "growing pains of a security system that's getting stronger, not weaker."
His reasoning goes like this. Many of these flaws have existed for a long time. AI tools, he said, are now surfacing vulnerabilities at a speed and scale never seen before, and he named tools like Mythos as examples. Enterprises are rushing to fix problems before bad actors can exploit them. "That surge in patching is disrupting things in the short term," he said.
DeWalt also sees a silver lining. He would rather see U.S. enterprises find and fix their flaws, even in a messy way, than have attackers find and exploit them first. Senator Mark Warner, vice chair of the Senate Intel Committee, called the disruptions a wake-up call and wrote on X that AI is moving faster than our ability to manage its consequences. TechCrunch notes that investigations are still underway.
The legacy hardware explanation
A second expert offers a different angle. Anusha Iyer, a software engineer and founder of the security startup Corsha, told TechCrunch that this may be the first moment when legacy technology meets modern technology at scale. She said industrial systems were never designed to be connected, let alone securely connected the way organizations are now pushing them to be.
She pointed to the FAA outage as an example. According to TechCrunch, preliminary findings suggest the FAA was modernizing its telecommunications infrastructure, and the transition overwhelmed parts of its older hardware. The new technology likely ran at normal modern capacity, while the legacy equipment could not keep up.
Iyer argues AI is getting an unfair reputation here. In her words, the problem is not rapid AI adoption but "the rushed mix of the legacy and modern systems." She says the missing pieces are a clear understanding of operational and cyber risk, and the ability to protect the older systems.
The two explanations are not identical, but they overlap. Both describe change happening faster than organizations can safely absorb it.
What this means for a small organization
You probably do not run air traffic systems, but the pattern applies at any size. Most small businesses, nonprofits, and public sector teams have an older server, a aging line-of-business application, or a website plugin nobody wants to touch. At the same time, they are being urged to add new tools, including AI features, on top.
The lesson from both experts is not to avoid change. It is to change in a controlled way. DeWalt said organizations need a steadier approach to AI adoption, including more testing and clearer priorities about what gets fixed and when. Iyer said older systems should be segmented more tightly and put under tighter access controls, so new systems can be phased in with less risk.
It is also worth remembering that many of the affected services are ones your team likely depends on every day, such as email, chat, and payments. When one goes down for hours or days, the question is whether your work can continue.
Practical steps to reduce outage risk
These are general good practices that follow from what the experts describe:
- List your older systems. Write down which servers, applications, and devices are old, who depends on them, and who maintains them.
- Test before you patch or connect. Try updates and new integrations in a test copy first, and roll them out in phases instead of all at once.
- Prioritize fixes. Not every flaw is equally urgent. Decide which systems are exposed or critical and handle those first.
- Limit access to legacy systems. Separate older equipment from the rest of your network where you can, and review who has access to it.
- Plan for downtime. Decide how your team will communicate and take payments if email, chat, or a payment service is unavailable for a day or more.
- Slow down on new connections. Before linking an older system to a new tool, ask what could break and who would notice.
If you are weighing where AI tools belong in your operations without adding risk, a plain-English AI roadmap can help you sequence the work.
What we still do not know
The causes of these outages are not fully settled. Officials and companies describe them as separate technical events, and investigations are ongoing. DeWalt's and Iyer's views are expert opinions, not confirmed findings, and they disagree on how much AI itself is to blame. What both agree on is that rushed change to complex, older systems carries real risk.
DeWalt expects more companies to focus on AI-driven vulnerability management, safe patch testing and deployment, and resilience for critical infrastructure. For now, the practical takeaway is simple: keep patching, but test first, and know which of your systems are fragile.
Sources
This post was drafted with AI from the reporting linked above and published by The Jones Foundry. For full details, read the original sources.