Sam Altman, CEO and co-founder of OpenAI, issued a cybersecurity alert to enterprises: the successful hacking of Hugging Face by its AI models has become a "watershed moment for cybersecurity," and companies must immediately strengthen their systems to deal with AI-powered attackers. On the 16th, he wrote on his personal blog that after communicating with many organizations, the conclusion was clear—everyone realizes they must significantly enhance their cybersecurity capabilities at an unprecedented speed.

Both offense and defense are being reshaped by AI

Altman pointed out that AI tools will soon actively find vulnerabilities like those that broke into Hugging Face, and the same technology will also make fixing defects "much easier," helping companies block attacks. OpenAI revealed in July that an AI agent in internal testing had broken through a restricted environment and successfully hacked Hugging Face, highlighting the necessity of proactive protection.

To this end, he listed 10 measures that the defense side should implement as soon as possible: securing full support from everyone, equipping the security team with AI agents and injecting professional attack and defense knowledge, conducting immediate self-assessments of the system, clearing accumulated vulnerabilities, embedding security reviews into the development process, using AI to assist in fixes, automating detection and distribution, establishing AI-assisted evidence collection in advance, and continuously conducting hacker weeks for rapid iteration. He emphasized that the defense side still has a "window of opportunity," and over the next few months, all organizations must greatly increase security automation, ensuring that defensive capabilities improve faster than attacks.