360 Security Says AI Model Matches Mythos Hacking Power as U.S.-China Risk Fears Grow
Updated
Updated · WIRED · Jun 24
360 Security Says AI Model Matches Mythos Hacking Power as U.S.-China Risk Fears Grow
3 articles · Updated · WIRED · Jun 24
Summary
360 Security Technologies said this week it built an AI model with hacking capabilities on par with Anthropic’s Mythos, sharpening concerns that frontier models are becoming usable cyber weapons.
Conference discussions in Beijing highlighted why that matters: more agentic models can expose code flaws, automate social engineering and enable new attacks, even as researchers expect defenses to improve later.
Open-weight AI is a key pressure point because Chinese firms including Moonshot, Alibaba and Z.ai have pushed highly capable models, while experts warn even weaker open models could become dangerous if guardrails are removed.
Some Chinese developers are already pulling back from open-source releases over security worries, while researchers at the event argued U.S.-China cooperation on standards and risk reduction is increasingly necessary despite strategic rivalry.
As open-source AI learns to hack critical systems, how can a global cyber catastrophe be prevented?
China weaponizes AI for 'cognitive warfare.' Can a genuine US-China safety partnership ever be achieved?
The $12 Trillion AI Cybersecurity Arms Race: China’s 360 Security Challenges Anthropic’s Mythos Amid U.S. Export Controls
Overview
China's 360 Security Technology has launched the Yitian Tulong AI suite to directly counter Anthropic's Mythos model, which previously alarmed global cybersecurity experts for its advanced vulnerability discovery abilities. The suite, introduced at the ISC.AI 2026 conference, includes Tulongfeng for automated vulnerability detection and Yitianzhen for cyber defense and incident response. Framed as critical national assets, these tools reflect China's focus on digital sovereignty and security. This move highlights the escalating AI arms race, as China seeks to match or surpass U.S. capabilities and address concerns about one-way transparency in cybersecurity.