Who Found the PHP Bug That Made Governments Unplug Their Websites? ChatGPT, Of Course
OpenAI says its models helped discover two serious PHP vulnerabilities that allowed attackers to run code on affected servers. The flaws, including one believed to have existed for 14 years, were found by researchers working with OpenAI and independently by Wiz. PHP released fixes on October 21 after details leaked ahead of the planned disclosure date; government websites in Finland, the Netherlands and France were taken offline amid the response. OpenAI is using the case to promote AI-assisted cybersecurity, proposing that authorized defenders scan and, in extreme cases, remotely patch vulnerable servers. The company also pledged $1 billion in subsidized access to its Daybreak security program and says it is investigating other bugs, including a Node.js remote-code-execution flaw. The piece questions OpenAI’s framing, citing prior incidents involving its agents’ access to non-public data and systems.
