China Unleashes Open-Source AI to Challenge Mythos
China responds to U.S. export controls on Anthropic's cybersecurity AI with Qihoo 360's Tulong Feng and Z.ai's free MIT-licensed GLM-5.2, which outperforms Claude Code in vulnerability detection at a lower cost.
Quick Take
Qihoo 360 unveils Tulong Feng as China's domestic answer to Mythos.
Z.ai releases GLM-5.2 under MIT license, beating Claude Code on benchmarks.
U.S. restrictions on Anthropic spark China's open-source AI push.
Zhou Hongyi calls Mythos 'cyber nuclear weapons' inaccessible to China.
Market Impact Analysis
NeutralArticle focuses on AI and cybersecurity, with no direct implications for cryptocurrency markets.
Speculation Analysis
Key Takeaways
- China fired back at U.S. AI export controls with open-source cybersecurity models that rival Anthropic's Mythos.
- Z.ai’s GLM-5.2, released under MIT license, scored higher than Claude Code on vulnerability detection at $0.17 per finding.
- Qihoo 360 unveiled Tulong Feng, claiming 3,432 vulnerabilities found, 105 confirmed by Chinese regulators.
- The open-access push removes geographic restrictions, directly contrasting with Anthropic's vetted access program.
What Happened
Chinese tech firms rapid-fire launched two AI cybersecurity models after the U.S. pulled Anthropic's Mythos and Fable models from foreign access. Qihoo 360 founder Zhou Hongyi took the stage at ISC.AI 2026 in Beijing to unveil Tulong Feng, calling it China's answer to the restricted American system. Simultaneously, Beijing-based Z.ai (Zhipu AI) released GLM-5.2 under a permissive MIT license, making it freely available worldwide. The double launch marks a direct retaliation against U.S. export controls that locked Chinese developers out of Anthropic's cybersecurity tools.
The Numbers
In head-to-head tests, GLM-5.2 hit a 39% F1 score on insecure direct object reference detection, edging out Claude Code. The cost differential is stark: each vulnerability finding cost about $0.17 with GLM-5.2, compared to over $1 using Claude workflows. Meanwhile, Qihoo 360 claimed Tulong Feng has unearthed 3,432 vulnerabilities cumulatively, with 105 confirmed by Chinese regulatory bodies. The open-source MIT license for GLM-5.2 eliminates the paywalls and geographic restrictions that now surround Anthropic's tools.
Why It Happened
The U.S. placed Anthropic's cybersecurity AI models under export controls, restricting access to a vetted coalition that included Microsoft and Apple but excluded Chinese companies. Zhou framed Mythos as “cyber nuclear weapons” that Chinese organizations couldn't even inspect. The exclusion galvanized a domestic AI push: China accelerated development of alternatives that could operate without U.S. dependencies. Open-sourcing the models became a strategic move to undercut the restricted access and build global adoption.
Broader Impact
The AI arms race enters a new phase as open-source models challenge proprietary, vetted systems. By releasing GLM-5.2 under MIT, Z.ai forces the question: can restricted access survive when comparable capabilities are freely shared? The move could fragment global AI collaboration but also pressure U.S. firms to rethink licensing strategies. Expect more open-source cybersecurity AI from China and possibly other nations seeking self-reliance.
What to Watch Next
- Whether Anthropic regains export clearance for Mythos and Fable models as negotiations with Commerce Department continue.
- Adoption rates for GLM-5.2 and Tulong Feng among global cybersecurity firms — will price and accessibility win out?
- Potential U.S. countermeasures, including expanded restrictions or incentives for American AI development.
This article is for informational purposes only and does not constitute financial advice.
Always late to trends?
Join for the latest news, insights & more.
Disclaimer: Bytewit is an independent media outlet that delivers news, research, and data.
© 2026 Bytewit. All Rights Reserved. This article is for informational purposes only.
