Homeland Security chief relays unverified AI theory for recent outages
Microsoft, Chase and the FAA offered software or network explanations. Mullin’s televised remarks were not a finding that AI caused the disruptions.
Loading page…
Microsoft, Chase and the FAA offered software or network explanations. Mullin’s televised remarks were not a finding that AI caused the disruptions.
Listen to this story
The outages now sit between two very different accounts: Homeland Security Secretary Markwayne Mullin relayed an unnamed contact’s theory that rapidly self-modifying AI software could outpace surrounding systems, while operators described separate software and infrastructure faults. Mullin said he was 98% certain the incidents were not cyberattacks, but his AI explanation is unverified—and Microsoft explicitly ruled out AI agents in its outage. The practical question is whether ongoing investigations, especially into the FAA disruption, produce evidence connecting these failures or leave them as unrelated incidents.
Microsoft’s nearly 24-hour November 12 outage affected Outlook, Teams and Exchange Online; the company cited a July software bug, high traffic and reduced server capacity.
The November 13 ground stop affected radar feeds and communications at 17 of 22 FAA air-route traffic control centers; the root cause remains under investigation.
Chase attributed its November 16 disruption to a software bug in hardware it operates on its premises; it reported no evidence of a cyber link.
Homeland Security Secretary Markwayne Mullin brought an unnamed technology contact’s theory about AI into the debate over outages affecting email, flights and banking. Speaking on Fox News on November 16, he suggested AI might be involved, while calling the incidents individually unrelated and saying he was 98% certain they did not involve a cyberattack, according to Fortune.
Mullin told host Sean Hannity that his contact suspected programs that write code and fix themselves could overwhelm systems unable to keep pace. He used the term “super intelligence” for the technology and specifically said his source was not Elon Musk. The proposed mechanism concerned software changing faster than surrounding systems could handle—not an identified attacker breaking into those systems.
That explanation remained his contact’s theory, not an established cause of the outages. Mullin also compared changes in technical systems to work on a house, where changing one thing creates a ripple effect elsewhere. His description allowed for disruption without a deliberate attack, even as he reassured viewers about the incidents.
with 98% certainty, we’re not seeing a cyber attack
Markwayne Mullin, speaking on Fox News, as quoted by Fortune
The disruptions unfolded across several days. Microsoft’s November 12 outage affected Outlook, Teams and Exchange Online. A nationwide flight ground stop followed on November 13, and Chase experienced hours of banking disruption on November 16. Each operator offered a different technical explanation:
The official reassurance has not satisfied every security expert. Marcus Rogers, director of Purdue Polytechnic Institute’s Cyber Forensics Lab, told the Los Angeles Times that unexplained failures deserve investigation even if they are not attacks. He challenged companies that say they do not know a disruption’s cause while confidently ruling out a cyberattack. His concern was the basis for that certainty, not proof of AI involvement.
AI critic and NYU professor emeritus Gary Marcus shared the unnamed contact’s intuition, but stopped short of claiming it had been proved. He told Fortune that outside researchers might examine public logs for evidence. Asked which of the AI risks he described explained the recent outages, Marcus said: “We don’t really know yet.” His proposed route forward was investigation rather than a definitive diagnosis.
At the FAA, a technical account has not yet settled the underlying cause. Transportation Secretary Sean Duffy said four days after the ground stop that the root cause remained unclear. He separately said the agency’s SMART artificial intelligence tool played no role.
Story updates
Recent outages have renewed concern that hurried repairs to vulnerabilities found by AI can disrupt systems. But the incidents are not established as connected, and JPMorgan Chase explicitly denies AI was involved in its outage.
Efforts to fix flaws found by AI are being blamed for some recent outages—not attacks by AI. A spokesperson for Senator Mark Warner’s team linked botched patch implementation to NSA and banking disruptions in comments published by Fortune on November 18. But JPMorgan Chase rejected an AI connection to its own outage.
The Washington Sun reported on November 17, citing people familiar with the matter, that NSA and Wall Street executives had told members of Congress that efforts to repair AI-discovered vulnerabilities caused recent outages, according to Gizmodo’s account. The claim concerns defensive work intended to prevent breaches, rather than an attacker using AI to disable services.
A congressional official quoted by the Sun described organizations trying to push through more patches at once than their systems were designed to handle. AI models were finding serious vulnerabilities, the official said, increasing the pressure to repair them immediately. Futurism also carried that account; its congressional claims trace back to the same Sun reporting.
The NSA disruption has a more specific public explanation. Director of National Intelligence Jay Clayton told Fox Business on November 17 that patches for recently discovered PHP software vulnerabilities caused the outage affecting its classified internal communications network. Reuters reported that AI had discovered those vulnerabilities, according to Gizmodo.
Bank of America, Wells Fargo and JPMorgan Chase all suffered outages during the week. On November 16, banking customers reported trouble accessing accounts and transferring money. JPMorgan attributed its transaction delays to a software bug in its internal systems.
Its spokesperson gave Fortune an explicit denial of AI involvement: the disruption came from a temporary software bug in hardware systems running on the bank’s premises. OpenAI’s tools were not a factor, the spokesperson said. The bank also said it had neither identified that bug through AI tools nor used AI to fix it.
OpenAI offered a different explanation to Fortune, attributing JPMorgan’s outage to older systems being overwhelmed by code modernization work. That statement described pressure from changing software, while JPMorgan pointed to a particular hardware-system bug and expressly ruled out AI tools.
AI now discovers vulnerabilities at machine speed while our testing and deployment safeguards remain stuck at human speed.
The repair workload was already substantial at one affected bank. Bank of America disclosed in October that it had hired 1,000 contractors to help patch flaws exposed by Anthropic’s Mythos model, alongside 3,000 internal staffers, CNBC reported, according to Fortune. That staffing disclosure concerned the scale of patching work, not a finding about the cause of this week’s outage.
Loading discussion...
Amit Shah, founder and CEO of InstaLILY AI, called the problem a failure of implementation. He told Fortune that an AI tool can spot a vulnerability without knowing what else a repair might break. A person approving the recommendation may see the proposed fix without seeing all the systems it touches.
Cliff Steinhauer, director of information security and engagement at the National Cybersecurity Alliance, described a competing pressure. Teams do not always have time to test every update fully when fixes are numerous and attackers can exploit known vulnerabilities, he told Fortune. Unlike Shah, he viewed the disruption as a necessary phase of securing existing software.
He hopes discovery and patching will eventually peak, then slow as teams fix underlying flaws. He compared the tradeoff to closing a bridge to repair a crack: disruptive, but better than a collapse.
The incidents still cannot be treated as one connected failure. Homeland Security Secretary Markwayne Mullin told a House hearing on November 18 that officials had seen nothing indicating they were related. Warner separately said there was no indication of a cyberattack, foreign interference or a coordinated effort.
Join the conversation
Explain when early warning outweighs the risk of misleading people.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.