Anthropic says China’s GLM-5.3 nearly matches Mythos at cyber exploits

Anthropic's Frontier Red Team found that China's open-weight GLM-5.3 model nearly matches Anthropic's Mythos Preview in creating cyber exploits, bypassing safeguards with simple methods. GLM-5.3 generated working exploits in tests and discovered unknown browser flaws, highlighting increased cyber risks from accessible AI models.