A Chinese cybersecurity firm, 360 Digital Security Group, has claimed that its AI-driven vulnerability discovery system can achieve performance levels comparable to Anthropic’s newly unveiled Claude Mythos model. This assertion follows the firm’s notable success at the revived 2026 Tianfu Cup hacking competition, where its system reportedly helped identify nearly 1,000 vulnerabilities, including serious flaws across various platforms. Researchers, including Eugenio Benincasa from ETH Zurich, have noted that while 360’s capabilities are impressive, they may not yet match the advanced reasoning abilities ascribed to Claude Mythos. Furthermore, Benincasa highlights that Chinese regulations require private firms to report vulnerabilities to government entities, which may give China a strategic advantage in cybersecurity research compared to other nations.

Mozilla: Mozilla develops the open-source Firefox web browser and prioritizes proactive security measures. It gained early access to Anthropic’s Claude Mythos through Project Glasswing to scan for vulnerabilities ahead of Firefox releases. This collaboration demonstrated AI’s potential to identify flaws that align with elite human researcher capabilities.
Anthropic: Anthropic is an AI safety and research company developing advanced Claude models with strong emphasis on reliability and security applications. It recently launched Claude Mythos Preview, a frontier model capable of autonomously discovering and exploiting vulnerabilities in critical software, accessible only via Project Glasswing to a limited set of partners. The model has been praised for accelerating vulnerability hunting while raising concerns about potential misuse.
Eugenio Benincasa: Eugenio Benincasa is a senior researcher at ETH Zurich’s Center for Security Studies, specializing in Chinese cybersecurity trends and hacking ecosystems. In a recent Natto Thoughts blog post, he evaluated claims by 360 Digital Security Group, concluding their AI shows significant progress but falls short of Claude Mythos’ reasoning depth. He emphasized China’s legal framework directing such discoveries toward state intelligence.
Palo Alto Networks: Palo Alto Networks provides comprehensive cybersecurity solutions including network protection and cloud security platforms. As a Project Glasswing partner, it tested Anthropic’s Claude Mythos Preview and reported enhancements in automated vulnerability discovery processes. The firm views the model as signaling a shift toward continuous AI-driven defenses.
360 Security Technology: 360 Security Technology, operating under Qihoo 360, is a prominent Chinese cybersecurity company focused on vulnerability research and participation in elite hacking contests. Its 360 Digital Security Group developed a Multi-Agent Collaborative Vulnerability Discovery System that secured first place at the 2026 Tianfu Cup competition. The firm recently claimed this AI system discovered critical flaws comparable to those found by Anthropic’s Claude Mythos.

`json
{
“Project Glasswing”: “Anthropic’s Project Glasswing provides select organizations with access to Claude Mythos to enhance security for foundational software systems.”,
“Chinese Regulation”: “Chinese legislation mandates private cybersecurity firms to report vulnerabilities to government agencies before making them public, directing research into state channels.”,
“Tianfu Cup Revival”: “The Tianfu Cup hacking contest resumed under Ministry of Public Security oversight, emphasizing AI’s role in competitive vulnerability discovery.”
}
`