Chinese cybersecurity leader 360 Digital Security Group has announced that its AI-driven "Vulnerability Discovery Agent" has identified nearly 1,000 previously unknown software vulnerabilities. These flaws affect widely-used products including Microsoft Office and the open-source AI framework OpenClaw, marking a significant leap in AI's role from a supporting tool to a core component in threat detection.
AI-Powered Discovery Advances
According to a report by Eugenio Benincasa from ETH Zurich's Center for Security Research, 360's agent leverages deep learning models to automate code scanning and pattern recognition, achieving far higher efficiency than traditional methods. The system can cover more code paths and reduce false positives, enabling faster patch cycles and proactive defense for enterprise clients.
Direct Competition with Anthropic's Mythos
The report highlights that 360 is positioning this agent as a direct competitor to Anthropic's Mythos, a leading AI security system focused on vulnerability prediction and remediation. Earlier this year, 360 also introduced a separate AI tool designed to accelerate the creation of exploit chains — critical steps in full system compromise. This dual-pronged strategy shows 360's ambition to own the entire AI security toolchain from discovery to exploitation.
Industry Impact and Outlook
The discovered vulnerabilities span everyday office software to foundational AI infrastructure, posing potential risks of large-scale data breaches or system takeovers if exploited. 360 stated it has responsibly disclosed all bugs to vendors and assisted with patches. Analysts warn that while AI-driven defense is advancing, attackers could equally leverage similar tools to find zero-days faster, intensifying the cyber arms race. The development comes amid broader industry moves: OpenAI's "Daybreak" project and U.S. VP JD Vance's calls for addressing AI threats to critical infrastructure. The AI security landscape is poised for rapid transformation.