China Has Matched Anthropic in Cybersecurity, Resetting AI Race

Live updates

1. -3116s ago

   China Matches Anthropic in Cybersecurity Capabilities

   Chinese AI systems have reached parity with Anthropic's Mythos in detecting software vulnerabilities. Tensions are rising as the U.S. targets Moonshot AI over alleged theft of Anthropic Fable. Beijing has labeled U.S. sanctions threats as "AI hegemonism."

   What's confirmed:

   • Beijing's MOFCOM called U.S. sanctions threats "AI hegemonism."
   • A congressional tabletop exercise used an AI agent called Red Kraken to simulate Beijing's cyber strategy.

   Still unconfirmed:

   • Moonshot AI's Kimi K3 was developed through distillation of Anthropic Fable.
   • China could use AI agents to target ports and logistics to slow U.S. mobilization.
   • Researchers dispute the 15-day Fable timeline.
   confidence 80%
2. 24d ago

   Chinese AI Models Match Anthropic's Mythos in Cybersecurity

   Chinese AI systems have reportedly reached parity with Anthropic's Mythos in identifying software vulnerabilities. This development comes amid U.S. administration restrictions on OpenAI and Anthropic. The shift is raising national security concerns in the United States.

   What's confirmed:

   • Chinese AI models have matched Anthropic's Mythos in finding software vulnerabilities.
   • Zhipu AI's open-weight GLM-5.2 and 360 Security Technology's Tulongfeng are among the models that have matched Mythos.
   • The Trump administration limited access to OpenAI's GPT-5.6.
   • Anthropic's general-use model was shut down for two weeks before Mythos was partially restored.

   Still unconfirmed:

   • China has developed an AI cyber nuclear weapon.
   • A Chinese cybersecurity company has built a bug finder that is better than Mythos.
   • Z.ai can match Mythos on cybersecurity.
   confidence 90%