One Week, Six Outages: AI's Security Sweep Is Breaking What It Protects
Six outages in one week hit Microsoft, JPMorgan, Apple, the FAA and the NSA. Officials say no attack — but AI-driven patching may be the common thread.
By Nathan Brooks
4 min read

What's News
- The FAA's Friday ground stop lasted over four hours — the longest nationwide stop in 25 years.
- Microsoft attributed its global outage to a 'latent defect' in a network fix deployed since summer.
- The NSA's classified communications network went offline Tuesday during scheduled maintenance.
- DNI Jay Clayton linked the NSA outage to patching after an AI-detected PHP vulnerability alert.
- U.S. cyber authorities issued an advisory Wednesday urging installation of Oracle's Java security update.
A single week knocked out Microsoft 365, grounded every U.S. flight for four hours, delayed JPMorgan Chase transactions and took down Apple Music, Marriott's apps and even the NSA's classified communications network. The Trump administration said there is "zero" indication the outages are linked or caused by bad actors like cybercriminals — and the more uncomfortable explanation may be the opposite: some of these failures came from the effort to make systems safer, with AI now sitting on both sides of that equation.
What actually went down?
The failures piled up fast and across sectors that share almost nothing except interconnection:
- Microsoft suffered a global outage late last week that disrupted Outlook, Teams and the AI assistant Copilot. The company traced it to a "latent defect" in a network fix it had begun deploying over the summer.
- The FAA's network malfunction triggered the longest nationwide ground stop in 25 years on Friday, lasting more than four hours and forcing carriers to cancel and delay swaths of flights.
- United Airlines delayed departing flights Monday after its own system outage.
- JPMorgan Chase said a software bug hit its internal hardware systems Monday, causing transactional delays and sending jitters through the financial system before the problem was resolved.
- Apple on Tuesday briefly disrupted Apple Music, the App Store and AppleCare after what it called an internal update; Marriott reported sign-in failures on its website and app.
- The NSA said an internal network used to communicate classified information was offline Tuesday "during scheduled routine maintenance."
The government, Microsoft, United, JPMorgan, Apple and Marriott all maintain there's no evidence the disruptions are linked or the result of a cyberattack. Microsoft also stated its outage did not stem from "AI / agents," though it didn't specify the goal of the summer network fix. It did not respond to an Associated Press request for details, and the other companies offered no additional comments Wednesday.
What does this change for AI and security teams?
Here's the twist practitioners should sit with: AI is both finding the bugs and straining the systems. AI recently helped detect sweeping vulnerabilities in PHP, a scripting language that powers websites worldwide, triggering urgent update calls — and that patching work itself caused limited outages and service delays. Director of National Intelligence Jay Clayton told Fox News the post-PHP patching has caused some systems to "not come back up" as expected. "That is what has happened at the NSA today," he said.
Cybersecurity startup Calif said it used AI to uncover a software bug in Java, one of the world's most widely used programming languages. Oracle, which owns Java, released a security update, and U.S. cyber authorities issued an advisory Wednesday urging organizations to install it.
The risk side is growing too. "As AI adoption accelerates, we are likely to witness more incidents of this nature," said Raj Rajarajan, director of the Institute for Cyber Security at City St George's, University of London. He noted much of today's digital infrastructure wasn't designed to handle the "computational demands of large-scale AI services."
NYU Tandon cybersecurity professor Justin Cappos put it more bluntly: "AI is absolutely making it way easier for attackers to go and launch cyber attacks. It is not quite as good at creating patches for bugs, like fixing problems, as it is exploiting them."
A Google cybersecurity report this week said hackers are "moving beyond basic prompting," using large language models and "agentic frameworks" to reverse-engineer security patches and exploit the vulnerabilities they reveal.
The concentration problem
Content creator Jake Smith, who posts political videos on Instagram and YouTube, captured the public mood: "To see several disruptions this close together, I think to a lot of people it's made them realize how many parts of their lives they've handed over to these systems that we still, at this point, don't fully understand. I think it gives people a sense of helplessness."
Sen. Mark Warner, co-chair of the Senate Cybersecurity Caucus, called the disruptions "a wake-up call" and said "AI is moving faster than our ability to manage its consequences," while criticizing the Trump administration's lack of urgency. Still, he conceded that "based on what we know today, there is good reason to believe these disruptions are the unintended consequences of efforts to make our systems safer."
For IT and security teams, the practical takeaway is unglamorous but real: patching cycles built around AI-discovered vulnerabilities now carry their own outage risk, and scheduled maintenance windows deserve the same failure planning as the bugs they fix.
Source: Fast Company
More from Nathan Brooks
Show full bio
News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.
2,460 articles