Critical PHP vulnerabilities trigger global website shutdowns; Node.js and LiteLLM also compromised

Two critical vulnerabilities in PHP, a server-side programming language used by approximately 70% of websites worldwide, were discovered on 21 October 2026 using artificial intelligence tools developed by OpenAI and security firm Wiz. The flaws allow unauthenticated remote code execution, potentially giving attackers direct access to server file systems and databases. The PHP Foundation released patches on Wednesday afternoon, but the severity of the threat prompted governments and organizations globally to take precautionary action.

Government websites across multiple countries went offline ahead of the patch release. Finland temporarily shut down its government public web services, while the Netherlands took offline its National Institute for Public Health and the Environment website and the Tweede Kamer parliamentary website. Japan's Ministry of Health, Labor and Welfare suspended its information disclosure system, and the BBC's websites for programmes including Doctor Who and The Traitors became inaccessible. The Wikimedia Foundation, which operates Wikipedia, took down systems as a precaution. Cybersecurity agencies in the UK, EU, Australia, Japan, and the US advised organizations unable to apply patches immediately to take affected systems offline until remediation was complete.

The vulnerabilities had remained undetected in all PHP versions since version 5 (released in 2004), making them decades-old flaws. Cybersecurity experts noted that before AI-powered discovery, such flaws were difficult for attackers to exploit. However, AI tools can now adapt exploits across different PHP versions, configurations, and memory layouts, fundamentally changing the threat landscape. Early attacks were reported emerging after disclosure, prompting Cloudflare and other infrastructure providers to update firewall rules and mitigations.

The PHP crisis coincided with additional cybersecurity incidents on 21 October. Node.js released patches for four high-severity vulnerabilities, including one permitting remote code execution and two allowing TLS validation bypass. LiteLLM, an open-source proxy for routing requests to multiple large language model providers, was compromised when an attacker injected malicious code into releases published to PyPI, Docker Hub, and GitHub Container Registry. The malicious code included an information stealer targeting cloud credentials and AI service credentials, plus a remote access trojan. Additionally, Indian certificate authority eMudhra reported compromised private keys, potentially allowing fraudulent certificate issuance and website impersonation.

Cybersecurity professionals characterized the combined incidents as a major cleanup operation. The challenge extended beyond simple patching: organizations needed to identify every instance of PHP running across public websites, content management systems like WordPress, Drupal, Joomla, and Magento, shared hosting platforms, and third-party products often invisible to system administrators. The White House stated that federal agencies were working with leading AI laboratories to accelerate fixes. Japan's Prime Minister Sanae Takaichi urged the public to verify whether their websites used PHP.