Last week saw a major Microsoft outage, with consequences spilling over into key systems in the United States. Due to a technical fault, all flights in the US were grounded for hours, while problems simultaneously emerged at the National Security Agency (NSA) and in the operations of the largest American banks.
A network upgrade at Microsoft caused overloading and server shutdowns, while upgrades to an ageing legacy network grounded thousands of aircraft over the weekend. In the incident involving US air traffic control, the problem arose while engineers were replacing outdated copper wires with fibre optics. Issues with the NSA's internal communication system, as well as Apple's iCloud and Apple Pay, were attributed to things going wrong during system updates. The payment disruption in American banks was attributed to a 'software error' at JP Morgan Chase, which claims it was not a cyber-security incident.
A major flaw was also discovered in a popular programming language. AI was used to find vulnerabilities in Java, the third most popular programming language in the world, which is led by Oracle and used by almost all the largest companies on Earth. Researcher Scott Helme discovered malicious code in Optimizely, a platform used by some of the world's largest companies. Hackers managed to steal around $7,000 (£5,300) in cryptocurrency before they were detected and removed. Helme told BBC Radio 4's Today programme: 'I would be very surprised if all these cyber-attacks and the attackers behind them were not using AI to speed things up.'
Since the summer, AI companies such as Anthropic and OpenAI have decided it is too risky to give the wider public access to their most advanced models before allowing certain companies and institutions to use them. Anthropic, among other leading AI firms, has warned of the risks posed by its own products. Security experts can use the same technology to find vulnerabilities themselves and warn others before criminals get to them. This happened in the recent Java incident, as well as the one involving PHP, where there was a rush to patch vulnerabilities before they could be exploited. In cases like PHP and Java, the 'good guys' found the flaws, with the help of AI, before the 'bad guys', enabling them to prepare fixes before the errors could be widely exploited.
The world has stopped in recent years as CrowdStrike, and then Amazon Web Services, halted many devices and systems. Some of the outages can be attributed to technical debt - a term used when operators choose quick software fixes instead of addressing the underlying problem. The accumulation of superficial fixes can eventually lead to more serious issues. As AI prompts many companies to reassess their cyber-security, they may issue more updates to patch vulnerabilities they have found.
Outages are often a reminder of the physical technology on which our online world relies - such as cables, servers and data centres. They particularly affect the physical world because they often cause disruptions in healthcare, schools and people's ability to do their jobs. They also highlight the world's heavy dependence on a small number of actors. Infrastructure service providers such as Microsoft and Amazon, together with Google, run a huge portion of the world's computer servers. In the PHP incident, the world held its breath while one programmer - working voluntarily - worked on issuing a patch in an easily applicable form.
US Secretary of State Marco Rubio claims that recent incidents are not connected nor do they come from a single attacker. 'We are not under a systemic attack,' Rubio stated. 'These events, one has nothing to do with the other, are not interconnected.'
Sheila Pancholi, head of risk assurance at consultancy RSM UK, stated: 'Oracle's vulnerability in Java is a stark reminder of how dependent companies have become on a relatively small number of critical technologies.' She added: 'Organisations must know where their critical technology dependencies lie, act quickly when vulnerabilities appear and test whether their response actually worked.'
Xander Van Raemdonck, a security engineer at Aikido Security, stated that the challenge is 'knowing whether you are exposed'. 'Java is embedded in a huge amount of business software, so many companies do not have a complete list of where it is running,' Van Raemdonck said. AI tools used to find security flaws are still so new that they will likely continue to find vulnerabilities that have been lurking in code for years.
Martin Reynolds, field CTO at software delivery firm Harness, stated: 'AI accelerates discovery on both sides and accelerates how much software is delivered.' 'So, the amount of what developers need to patch is growing, while the time they have to do the job is shrinking,' Reynolds concluded.













