Cloudflare CEO Matthew Prince has warned that artificial intelligence could unleash years of serious cybersecurity threats, with newly discovered software vulnerabilities potentially causing widespread internet outages and data breaches.

The warning follows the discovery of major security flaws in PHP, a programming language used by an estimated 70 per cent of websites worldwide.

The vulnerabilities caused widespread disruption last week, forcing government websites, gaming platform GOG and numerous smaller websites offline while administrators rushed to apply security patches.

Prince believes the incident could be an early indication of a much bigger problem.

Speaking to Reuters, he revealed that Cloudflare is already aware of several other serious vulnerabilities affecting critical software, although he declined to identify the programs involved.

The growing concern is that increasingly powerful AI models are making it significantly easier for both security researchers and cybercriminals to uncover weaknesses in widely used software.

Previously, finding and exploiting complex security flaws required considerable technical expertise and weeks of painstaking code analysis.

Now, AI-powered tools can scan enormous codebases, identify potential vulnerabilities and help develop working exploits in a matter of hours.

A single researcher can deploy multiple AI agents to examine software simultaneously, dramatically accelerating a process that once required teams of experienced specialists.

Cybersecurity company Wiz Research, which helped identify the recent PHP vulnerabilities, confirmed that advanced AI models were instrumental in its investigation.

The company said the technology allowed researchers to develop and adapt working exploits across multiple applications within hours.

Wiz warned that software complexity may no longer provide the same level of protection against attacks, as AI reduces the expertise needed to identify and exploit hidden weaknesses.

The Electronic Frontier Foundation (EFF) has also sounded the alarm over the accelerating pace of vulnerability discovery.

According to the organisation, around three per cent of all software vulnerabilities ever recorded were identified in October alone, while nearly a quarter of known security flaws have been discovered during 2026.

The EFF believes advanced AI systems developed by companies including OpenAI and Anthropic are helping drive the surge by making previously difficult security research faster and more accessible.

Several vulnerabilities capable of potentially disrupting major parts of the internet have reportedly been discovered and patched in recent weeks.

Many affect open-source software that forms the foundation of websites, online services and critical digital infrastructure.

Although software vulnerabilities are nothing new, the cost and technical difficulty of discovering them previously acted as a barrier to widespread exploitation.

AI is rapidly eroding that barrier, potentially giving attackers access to weaknesses that might otherwise have remained undiscovered for years.

Both Prince and the EFF expect cybersecurity conditions to worsen before they improve, with more frequent attacks, data breaches and temporary service outages possible as additional flaws emerge.

For consumers, the EFF recommends keeping browsers, operating systems and applications updated with the latest security patches.

It also advises using password managers to create unique passwords for individual accounts, installing privacy tools such as Privacy Badger and remaining cautious of suspicious emails and phishing attempts.

Website owners should also ensure that the latest PHP security updates have been installed, either directly or through their hosting providers.

Despite the growing risks, Prince believes the wave of AI-assisted vulnerability discoveries could ultimately make the internet safer.

By exposing previously hidden security weaknesses, AI gives developers an opportunity to repair vulnerabilities before they can be exploited in future attacks.

However, the immediate challenge will be managing the disruption as researchers and hackers increasingly use the same powerful tools to search for flaws in the software underpinning the internet.