AI Uncovers Critical PHP Flaw Threatening Most Websites

Security researchers just found a massive hole in the code running most websites. They used artificial intelligence to spot it. The flaw sits in PHP, a language powering about seventy percent of the web. Experts call it php2shell. It lets attackers hijack or crash systems by messing with memory during queries. The National Security Agency got so worried they told unfixed sites to go dark for a bit. They feared hackers would use AI to exploit it fast.

A firm named Calif discovered the bug while testing AI tools from OpenAI. They started looking in August and found it within days. Surprisingly, many had already scanned PHP with AI and missed it. Calif reported it quietly so a fix could be ready before public disclosure. Cloudflare, a huge security provider, rolled out shields immediately. WordPress, another giant, screamed at site owners to update now.

This case shows a scary shift. Powerful AI models used to be locked down. Only vetted pros got access. Now Chinese models nearly as good are free for anyone. Amateurs can wield them. Earlier this month a South Korean bank got robbed by a China-based newbie using AI. The thief stole detailed customer data. Analysts say the attacker lacked skill but the tools did the heavy lifting.

OpenAI admits this is a double-edged sword. Their statement says AI can find internet-scale flaws but also helps bad guys build attack weapons faster. They hinted they've found more big vulnerabilities still being patched. The race is on. Defenders use AI to plug holes. Attackers use it to find new ones. The battlefield just got way faster and way more crowded.