AI is finding new holes in the infrastructure of the internet. Is anyone around to patch them?
For the past week, software engineers around the globe have been scrambling to update the backend of websites after a team using an OpenAI model discovered two major vulnerabilities in PHP, a critical piece of web software.
PHP, open-source scripting software that underpins as many as 70% of websites worldwide, can become outdated when website owners don’t update to newer versions of the software, making them vulnerable to attacks. Shortly after alerts were issued about the new vulnerabilities, new versions of PHP were issued, and site operators were urged to update. Some government websites went dark in Europe as these updates were applied, but for much of the web, it’s been left to individual website owners to apply the fixes.
No law compels them to do so, although if a website suffers a data breach, the owner could be liable to customers or clients. And for web surfers, it’s not obvious what PHP version a site is running without running terminal commands or using an online PHP checker.
Given that some websites are no longer actively managed — I own a website that was latent and hadn’t had a PHP update or updates of its WordPress software in years, for instance — the PHP story has shown that the web’s aging architecture is proving to be insecure due to neglect and dwindling resources to keep it safe.
One of the most shocking revelations that has come out of the last week’s race to get websites updated and safe is that an organization tasked with making sure security flaws in PHP are addressed is itself on shaky ground.
Source link







