Rubio says the recent incidents are not connected

A series of separate technology failures has disrupted Microsoft services, the National Security Agency’s internal communications, Apple’s iCloud and Apple Pay, and payments at major U.S. banks. An apparent technology glitch also grounded all U.S. flights for hours. The close timing prompted speculation that the incidents might be linked, but they arose across different companies and systems. BBC News said outages and IT failures are commonplace, but not on this scale.

U.S. Secretary of State Marco Rubio said the recent incidents were not connected or the work of one attacker. “We are not under systemic attack,” he said. “These events, one has nothing to do with the other, they are not inter-related.” The BBC report likewise said there was no evidence to suggest that the incidents were connected.

Many of the outages resulted from an update or installation going wrong somewhere in a system. A network upgrade caused Microsoft servers to become overloaded and shut down. Upgrades to a “creaky old network” grounded thousands of U.S. planes over the weekend, while the air traffic control failure came as engineers were replacing outdated copper wires with fibre. Problems with the National Security Agency’s internal communications system and, separately, Apple’s iCloud and Apple Pay were also blamed on things going wrong during system updates.

JP Morgan Chase separately put the disruption to payments at major U.S. banks down to a “software bug” and said it was not a cybersecurity incident. The same run of outages served as a reminder of the physical technology behind online services, including cables, servers and data centres.

Some recent problems involved cybersecurity vulnerabilities rather than accidental outages. AI was used to discover flaws in Java, described as the world’s third-most popular programming language. Scott Helme also discovered malicious code in Optimizely, a platform used by some of the world’s largest companies. Hackers stole about $7,000 (£5,300) in cryptocurrency before they were discovered and removed.

The Java and PHP incidents showed why early detection matters. Security professionals can use AI to find flaws and alert others before criminals get to them; in the PHP case, there was a scramble to issue patches before the flaws could be exploited. Since the summer, AI companies including Anthropic and OpenAI have decided it is too risky to give the general public access to their most advanced models before allowing certain companies and institutions to use them. Anthropic and other leading AI firms have also warned about the risks their own products pose.

Helme told BBC Radio 4’s Today programme: “I would be very surprised if all of these cyber-attacks and the attackers behind them were not using AI to accelerate themselves.” His statement was an assessment rather than a finding that every cyber-attack discussed involved AI. The report’s wider point was that AI tools remain so new that they will probably continue finding vulnerabilities that have lingered in code for decades. As companies reassess cybersecurity in response to AI, they may also issue more updates to patch flaws, increasing both the patch workload and the potential for disruption.

Outages are not new. In recent years, CrowdStrike and then Amazon Web Services grounded many devices and systems. One possible explanation for recurring failures is technical debt, defined in the report as operators using quick fixes to software instead of solving an underlying problem. A buildup of surface-level fixes can eventually end up with more serious problems, while the upgrades needed to address the debt can cause further disruption.

Outages can disrupt healthcare, schools, people’s ability to do their jobs and, in the weekend flight-control case, planes’ ability to fly. The concentration extends beyond frontier AI companies: Microsoft, Amazon and Google run a huge chunk of the world’s computer servers, while Oracle’s Java is used by nearly all of the biggest companies on Earth. During the PHP response, one programmer working voluntarily worked to issue patches in a form that was easily applicable.

Sheila Pancholi, risk assurance lead at consultancy firm RSM UK, said: “The Oracle Java vulnerability is a stark reminder of how dependent businesses have become on a relatively small number of critical technologies,” and said: “Organisations need to know where their critical technology dependencies sit, act quickly when vulnerabilities emerge, and test that their response has actually worked.” Xander Van Raemdonck, a security engineer at Aikido Security, said the challenge was “knowing whether you’re exposed.” Van Raemdonck said Java is bundled inside a huge amount of business software, so many companies do not have a full list of where it runs.

Martin Reynolds, field CTO at software delivery company Harness, said: “AI is accelerating discovery on both sides and speeding up how much software gets shipped,” and added: “So the volume of what developers need to patch is going up, while the time they have to get the work done it is going down.”

Continued work on technical debt means more system upgrades and more potential disruption. The report said hackers pose a constant threat and continue to get more sophisticated. Attackers need to find only one vulnerability to exploit, whereas defenders have to protect against every potential hack. Helme said: “Attackers can move quickly and carefree, and experiment without risk of failure so they can naturally often move a fraction faster.” He added: “We’re currently observing that small delay between how fast the attackers can move and how fast the defenders can move.”

The cat-and-mouse contest between attackers and defenders is unlikely to end soon, and the report said the impact of AI means the stakes are getting higher. The individual incidents remain separate, but together they expose how update failures, technical debt and concentrated reliance on critical technologies can disrupt the physical world.