Technology

Here’s How AI Exposed Long-Hidden Web Flaws Hackers Raced To Exploit

Wikimedia Commons/Public/Jernej Furman from Slovenia, CC BY 2.0

Artificial intelligence tools helped uncover dangerous flaws in the software behind millions of websites, sparking a race to fix them as hackers rushed to strike.

The PHP Foundation fixed two serious security holes within its software Wednesday after months of delays caused, in part, by a surge of AI-generated security reports, a foundation spokesperson told the Daily Caller News Foundation. PHP, a website programming language, powers websites for governments and banks, among the many other organizations, according to a Wednesday PHP Foundation blog post.

The Cybersecurity and Infrastructure Security Agency encouraged organizations to update their websites immediately and lookout for suspicious online activity, according to a Wednesday alert.

These security flaws, called “PHP2Shell,” could let hackers remotely take over vulnerable websites, according to Wiz Research. PHP powers roughly 70 percent of websites across the globe.

Cloudflare, which handles traffic for millions of websites, saw thousands of attempts to exploit the flaws within the 10 hours leading up to their public disclosure, Cloudflare CEO Matthew Prince said on X Wednesday. That number climbed into the millions following the announcement.

‘Find Bad Vulnerabilities In PHP’

Wiz, in addition to researchers at Calif and OpenAI, identified the flaws between August and September, PHP’s blog post stated. Calif used OpenAI’s GPT-5.6 Sol to find one of the flaws, OpenAI said in its Wednesday blog post.

“Frontier AI helped us find and fix vulnerabilities that had remained hidden in the PHP codebase for more than a decade,” Wiz researchers wrote in a Wednesday blog post.

The same OpenAI model was among those involved in the July breach of AI startup Hugging Face.

AI tools buried the PHP Foundation’s small team in bug reports that contributed to months of delays for a fix, a foundation spokesperson told the Daily Caller News Foundation. There was an “ever-increasing volume of incoming security reports due to the availability of AI tools,” which was among the reasons for the holdup.

Software security company Chainguard used AI to find the same flaw in one night after learning of its existence, Chainguard CEO Dan Lorenc told the DCNF.

“[W]e were able to independently discover the same issue using our harness and a frontier [AI] model with a basic prompt of ‘find bad vulnerabilities in PHP,'” Lorenc said.

‘Anyway, it’s Wednesday’

Social media users tracked the fallout as governments and companies around the world pulled their sites offline.

“PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse,” one X user wrote in a breathless post summarizing the event.

“Anyway, it’s Wednesday,” the user concluded, after cataloging roughly a dozen other breaches and cyberattacks that occurred throughout the week. The post has received 4 million views.

“You’re all prompting wrong,” an X user wrote in response to Lorenc’s LinkedIn post on the matter.

In the LinkedIn post‘s comment section, a security engineer asked Lorenc, “‘Find bad vulnerabilities in PHP,’ that’s all it took?” Lorenc replied, “yes.”

“There is nothing to worry about,” another X user wrote about the scale of the vulnerability. “Nothing important is run on PHP or old software — except the water & electrical grid, a few government services like our missile defense systems and a handful of electrified fences at the zoo (for the lions, tigers and cobras).”

One AI company’s security leader warned the worst is still ahead.

“As bad as things have gotten in the past 4 days, they are going to get a lot worse before they get better,” Anthropic Deputy Chief Information Security Officer Jason Clinton wrote on X Thursday.

‘I’m Sorry’

The PHP Foundation criticized Chainguard for releasing a software fix too early.

Chainguard “unilaterally broke the embargo” over six hours before a scheduled software patch, the foundation said in its Wednesday blog post, referring to an agreement among security companies to release the secret fix all at once before hackers can study and exploit it.

“We were disappointed that Chainguard made the decisions they did despite an agreed-upon embargo, but we hope that in the future, trust can be rebuilt with them,” the foundation told the DCNF.

The foundation also admitted its security team’s slow decision-making process and scarce resources were at fault for the delayed fix, according to the foundation’s blog post.

Lorenc released the fix early because his team found the flaw so easily and details had already leaked online, he said in a statement shared with the DCNF.

“I’m sorry that my decision led to even more stress for them,” Lorenc said, referring to the PHP Foundation.

“This is unfortunately the new normal for the industry, and we’re going to need to work together as one team to have a chance at securing our critical infrastructure fast enough to stop attackers from causing real harm,” Lorenc said.