Openai's AI model plays key role in uncovering major php security vulnerabilities

OpenAI’s GPT-5.6-Sol model helped uncover severe PHP security flaws, dubbed 'PHP2SHELL,' which have disrupted major websites. The vulnerabilities, existing for 14 years, were addressed through collaboration with PHP maintainers, Wiz, and Cloudflare. A smaller version was found in 2024 but overlooked. The coalition planned to disclose the flaws on October 22, but details leaked early, prompting fixes a day earlier. OpenAI’s President, Greg Brockman, stressed the importance of the two-month preparation due to the complexity of the fixes. The vulnerabilities caused outages at Wikipedia, Asana, and the Internet Archive, with many sites posting maintenance notices. OpenAI also worked with U.S. government agencies, including the White House’s Super Intelligence Force, and acknowledged AI’s dual use. The company committed $1 billion to subsidize access to its security tools and released a tool for checking vulnerable PHP versions. Japan’s Prime Minister urged website owners to update systems following the disclosure.