OpenAI used advanced AI to discover serious PHP vulnerabilities and launched a global collaborative repair mechanism. OpenAI issued an announcement on October 21, revealing to the outside world the entire process of fixing a PHP vulnerability that was discovered by AI and coordinated across multinational institutions and enterprises.

OpenAI used advanced AI to discover serious PHP vulnerabilities and launched a global collaborative repair mechanism. OpenAI issued an announcement on October 21, revealing to the outside world the entire process of repairing a PHP vulnerability discovered by AI and collaboratively handled by multinational institutions and enterprises. This action was promoted by the "Patch the Planet" (patch the entire Internet) project initiated by OpenAI, which brought together software maintainers, security researchers, cloud service providers and government network security agencies on the same front, with only one goal - to suppress the risks of the global Internet before vulnerabilities are targeted by attackers. This time, two serious vulnerabilities in PHP were actively "sniffed out" by AI. Security company Calif first discovered a serious security flaw with the help of GPT-5.6Sol; security company Wiz independently discovered and reported two related vulnerabilities. PHP is an open source language that supports countless websites and network applications. It relies on web page requests, database operations, and backend business logic. Once there is a hole in the bottom layer, websites, cloud services, and enterprise systems may be involved. Both vulnerabilities are rated as critical and could allow an attacker to remotely execute code without authentication. The PHP development team has launched a repair version on October 21 and urges users to upgrade immediately. OpenAI calls the most precious breathing period in this incident the "Defender's Window" - that is, the time difference between the defender and the attacker before the attacker discovers the problem, develops a repair plan, and deploys protection. This vulnerability was discovered with AI assistance just before the attacker took action, giving the maintainers valuable defense preparation time. But OpenAI also reminded that as model capabilities continue to climb, the time difference between discovering vulnerabilities and writing attack tools is being compressed. The industry must shift from the fragmented vulnerability handling method in the past to a systematic mechanism that can quickly coordinate maintainers, service providers, enterprises and governments. To this end, OpenAI has developed a coordinated response framework for serious vulnerabilities, which is divided into five steps. In the first step, maintainers, researchers and key technology partners work together to verify vulnerabilities, assess impacts, develop and test fixes, and conduct adversarial testing with advanced models to confirm that the patches can really block attacks and follow the clues to find other variants that may be exploited; in the second step, leave enough preparation windows for affected organizations, feed technical information in stages according to their respective protection responsibilities, and rely on information sharing agreements and specialized coordination mechanisms to suppress leak risks; in the third step, work with government cybersecurity agencies to ensure that critical infrastructure operators receive timely alerts. This time OpenAI contacted the White House Super