Yuki Chen Joins Z.ai Amid Rising Concerns Over AI Cybersecurity Risks

Yuki Chen, a prominent Chinese cybersecurity researcher, has joined the leading artificial intelligence model company Z.ai as the AI sector in China confronts the growing impact of technology on global cyberspace.

Chen’s appointment comes amid increasing concerns regarding AI-assisted cyberattacks and autonomous agents operating beyond their intended instructions. This includes recent incidents involving Chinese tech giant Tencent Holdings and US start-ups OpenAI and Anthropic.

In a social media post announcing his transition to Z.ai, Chen expressed his enthusiasm, stating, “New journey, new mission! Thrilled and grateful for the opportunity to help build a safer cyberspace with the power of frontier AI.”

Z.ai revealed that Chen, also known by the alias “Guhe,” recently aided the company in identifying 23 cyber vulnerabilities within Microsoft’s Windows operating system using Z.ai’s flagship GLM models, with Microsoft acknowledging these findings.

The company noted, “[Chen’s] decades of vulnerability research have helped, and will keep helping, the community build a safer cyberspace with GLM.” However, the announcement did not specify the level of human involvement in detecting the vulnerabilities or the model’s contribution.

Chen is recognized as one of China’s most legendary civilian “hackers,” boasting over 15 years of experience in identifying vulnerabilities in both software and hardware systems, according to his public profiles. He previously led the vulnerability research institute at Qihoo 360, a major Chinese cybersecurity firm known for its antivirus solutions, before becoming an independent researcher.

As part of the global “white-hat” community, Chen investigates software weaknesses and reports them to vendors for remediation. His contributions have garnered recognition from Microsoft and fellow security researchers, including a top ranking on Microsoft’s annual “Most Valuable Security Researcher” list five times, in 2019 and annually from 2021 to 2024.

In 2024, the Center for Security Studies at ETH Zurich identified him as one of China’s leading “superstar hackers” in the civilian sector. He has received the prestigious “Pwnie Award” twice, an honor often referred to as the “Oscars of hacking.”

Chen’s recruitment by Z.ai coincides with the increasing capabilities of GLM, which, like other frontier AI models globally, is becoming adept at discovering and exploiting cyber vulnerabilities. In August, Z.ai announced that its latest flagship model, GLM-5.3, was among the best in the world for vulnerability discovery, significantly outperforming its predecessor, GLM-5.2.

Last month, US start-up Anthropic reported that GLM-5.3 exhibited similar cyber offense capabilities to its controversial Claude Mythos model, including the ability to autonomously develop end-to-end cyber exploits. Following this report, Chen announced on X that he would begin using GLM-5.3 to assist in his research.

A report last month from the US government ranked GLM-5.3 as the most capable publicly released open-weight model for cybersecurity, although it noted that the model was still approximately four months behind the US frontier in cyber capabilities.

China’s AI industry is increasingly focusing on mitigating the risks associated with frontier AI, with ByteDance, the owner of TikTok, announcing the establishment of a new department dedicated to AI safety last week.

Additionally, China and the US agreed to initiate a bilateral AI safety dialogue at last month’s leaders’ summit in Washington, with AI expected to be a significant topic during President Xi Jinping and President Donald Trump’s remaining meetings this year.

Source: SCMP Tech (Asia)