Daijiworld Media Network - Beijing
Beijing, Jun 29: Chinese artificial intelligence models are rapidly closing the gap with leading US AI systems, particularly in the field of cybersecurity, according to security researchers.
A new AI model developed by Chinese firm Zhipu AI has reportedly demonstrated capabilities comparable to Anthropic's advanced cybersecurity model Mythos, which is designed to identify software vulnerabilities.
Mythos, previewed in April, was developed to detect security flaws in software systems. However, experts have warned that such technology could also be misused to enable more sophisticated cyberattacks. Recently, the United States restricted the export of a less advanced Anthropic AI model over national security concerns.
Researchers said Zhipu AI's models are capable of identifying software bugs and vulnerabilities, though they still lag behind leading AI systems from Anthropic and OpenAI in several other areas, according to a report by The Wall Street Journal.
The adoption of Chinese AI models is increasing as companies look for ways to reduce rising technology costs. Companies including Microsoft are reportedly exploring options to host Chinese AI models as businesses seek cheaper alternatives.
"China is making sure that the gap becomes smaller and smaller over time," said Lior Div, chief executive officer of cybersecurity firm 7AI.
Experts said the growing ability of AI systems to discover software vulnerabilities has increased the urgency to use the same technology to fix security flaws before hackers exploit them. Without adequate safeguards, researchers warned of a possible "bugmageddon" scenario involving widespread cyber vulnerabilities.
Zhipu AI's GLM-5.2 is an open-weight AI model that can be downloaded and operated by users on their own hardware. It can also be modified without external supervision, giving developers greater control but raising concerns that malicious actors could use the technology privately.
The development comes amid rapid advancements by Chinese technology companies in AI-powered cybersecurity. Cybersecurity firm 360 Security Technology recently launched two AI-based cybersecurity tools under its "Yitian Tulong" platform.
One of the tools, Tulongfeng, is designed to automatically detect software vulnerabilities, while Yitianzhen focuses on cyber defence and incident response.
The company described Tulongfeng as China's counterpart to Anthropic's Mythos, highlighting the growing competition between Chinese and US AI firms in the cybersecurity space.