China Unveils Advanced AI Cyber Weapons, US Braces for Impact
Artificial intelligence technology is now being developed to the point where it can be used as an advanced weapon in the cyber world. Chinese cybersecurity company 360 Security Technology has announced that it has developed technology described as a competitor to Mythos, the artificial intelligence (AI) system owned by Anthropic of the United States.
The company considers the AI model’s sophistication to be a strategic cyber capability that the Bamboo Curtain country must possess.
For context, Mythos, which was introduced last April, is a system designed to detect software vulnerabilities. However, a number of cybersecurity experts have warned that the technology also has the potential to strengthen cyber attacks.
Recently, the US government even ordered Anthropic to halt exports of a lighter version of the programme on national security grounds.
At the ISC.AI 2026 cybersecurity conference in Beijing, 360 Security Technology founder Zhou Hongyi introduced two AI-based security tools under the name ‘Yitian Tulong’. The name is taken from a classic Chinese martial arts novel and means ‘Heavenly Sword and Dragon Sabre’.
Zhou explained that one of the tools, Tulongfeng, is designed to automatically find software vulnerabilities and described it as China’s version of Mythos. Meanwhile, the second system, named Yitianzhen, was developed to automate cyber defence and responses to security incidents.
‘Powerful weapons that can change the landscape of cyber attack and defence like this should not be owned only by the other side,’ Zhou said in his speech, according to a transcript published by 360.
According to him, China faces the risk of one-way transparency if US entities can use models like Mythos to scan critical software and systems, while Chinese companies do not have equivalent capabilities.
360 claims Tulongfeng has found 3,432 software vulnerabilities, including 105 that have been confirmed by Chinese authorities. However, Reuters could not independently verify the claim.
Hampered by US chip restrictions
Since 2022, the US has further tightened export controls on China’s access to advanced American-made chips. The move has made it difficult for domestic Chinese AI models to catch up with US rivals, including Anthropic, although the gap is said to have narrowed since last year.
The US government argues the restrictions are necessary because advanced chips could help the Chinese military enhance its capabilities with AI support.
Zhou said that, objectively, domestic models still have a basic capability gap of around 20%-30%.
‘China cannot wait until model capabilities fully catch up before starting vulnerability discovery, because we cannot afford to wait,’ he said.
Instead, Zhou said his company is taking an agent-based path, namely combining AI models with cybersecurity expertise, vulnerability databases, and automation tools. He claimed that only 360 has succeeded in implementing this approach, enabling Tulongfeng to have capabilities equivalent to Mythos.
‘If Mythos is a high-end chip, then what we have built is a complete machine that can operate stably, work 24 hours a day, and make fewer mistakes,’ he said.
‘If the US approach is to nurture a genius hacker, then 360’s approach is to organise a professional attack and defence team,’ he concluded.