Skip to content
Claude News

anthropic

China reaches parity with Anthropic in vulnerability research

Claude News

Security researchers report that the new Z.ai model from Zhipu AI, released this month, now matches top-tier American models in vulnerability detection. It has reached the performance level of Anthropic's Mythos in bug discovery tasks.

Z.ai still lags behind Anthropic and OpenAI products in other domains. However, cybersecurity was the specific area where American models held a clear lead, as previously demonstrated by Mythos testing in NSA systems.

According to Bloomberg, this parity creates pressure on the White House during its review of U.S. AI policy, including export restrictions on the Mythos 5 and Fable 5 models.

Related stories

  1. Semgrep: GLM-5.2 outperforms Claude Code in IDOR detection
  2. NSA testing finds Mythos vulnerable
  3. Safety research on Fable 5 and Opus 4.8 models
  4. Anthropic ties Moonshot traffic to Chinese military
  5. Anthropic will bill again for requests its safeguards block
  6. Anthropic's 225 bug finds, one attack in the wild

Comments

No comments yet. Be the first.

Join the conversation

Sign in with Google to leave a comment. Your name and avatar come from your Google profile, and the comment appears after moderation.

We only use your name and avatar from Google. We never store your email address.