Cloudflare head believes AI development will lead to weekly "crashes" of the entire internet for several years

If you spent much time online this week, you likely noticed how it briefly felt like the end of days as government websites, game marketplaces, and furry image boards alike were pulled offline in a scramble to address newly-discovered security vulnerabilities in PHP, a scripting language used by roughly 70% of the world's websites, according to some estimates.

It was a harrowing incident, revealing flaws in foundational internet infrastructure that might have turned some of the most popular web destinations into attack vectors.

Cloudflare CEO Matthew Prince said in an interview that we should expect similar episodes on a regular basis, as LLM technologies accelerate the discovery of exploitable security vulnerabilities.

Cloudflare is a major cloud networking and cybersecurity provider that a number of websites rely on. Prince said the company is aware of "a bunch of additional severe vulnerabilities that are present in other pieces of critical software" like those that triggered the PHP panic earlier this week. Both cybersecurity experts and bad actors are tasking LLMs with identifying previously unrecognized weaknesses in the core technologies of the internet, so more discoveries will likely follow.

I think we're going to have a couple years where every single week it's going to feel like the sky is falling.

Previously, the rate at which exploitable flaws could be identified was limited by the finite number of experts able and willing to devote the necessary time and attention. Likewise for those seeking to weaponize those flaws. For a user directing a workforce of LLM agents, those limitations are less relevant.

Wiz Research, one of the firms that identified the recent PHP exploits with the help of LLM tools, wrote in a blog post detailing its findings that "the economics of both discovery and exploitation" have changed.

Frontier models were critical to making this process practical, enabling us to develop and adapt working exploits across multiple applications in hours. As the time and expertise required for this work decrease, complexity may no longer provide a meaningful barrier, changing how we define what is vulnerable in practice.

Prince expects busy months ahead for those tasked with handling the rising tide of LLM-empowered exploits. In the long term, though, he believes software will become more secure as the newly-identified vulnerabilities get patched.

"The reality is, we're going to get a lot of this stuff cleaned up," he added.

Tags: