AI Could Surface the Internet's Hidden Fragility
Original: Will AI expose the rickety infrastructure of the internet?
Why This Matters
AI-accelerated exploit discovery threatens infrastructure built on aging open-source code.
A critical PHP vulnerability forced site owners to patch immediately or go offline, spotlighting how AI tools could accelerate discovery of flaws buried deep in the open-source foundations of the web.
A newly disclosed set of vulnerabilities in PHP—one of the most widely used open-source programming languages on the web—triggered urgent warnings this week: patch immediately or take your site offline. The incident, dubbed 'PHP-ocalypse' by some commentators, prompted governments to pull down websites and sent popular platforms into brief outages while scrambling to apply fixes.
On TechCrunch's Equity podcast, hosts Anthony Ha, Sean O'Kane, and Kirsten Korosec unpacked what this means for the broader state of internet infrastructure. O'Kane framed it bluntly: the web is 'held together with glue and dreams,' and moments like this expose just how fragile that foundation is. He noted that researchers from labs including Google and OpenAI were involved in surfacing vulnerabilities—though the specific PHP flaws were found by human researchers, not AI models.
The real concern, O'Kane argued, is the near future: as LLMs become more capable and AI agents can be tasked autonomously to probe systems, the pace of vulnerability discovery—by defenders and attackers alike—will accelerate sharply. Korosec raised the stakes further, warning that banks and financial institutions could be prime targets, and that the scale of potential exposure 'feels a lot bigger than we can fix right now.' The episode positions this week's events less as a one-off crisis and more as a preview of what systematic AI-assisted vulnerability hunting could look like at scale.