The Powerful Chinese Model Experts Warned About—and Waited for—Is Here

Details

Chinese AI company Z.ai released GLM 5.3, an open-weight model it claims performs cybersecurity and coding tasks nearly as well as top closed models from Anthropic and OpenAI. Unlike Claude or GPT-4o, GLM 5.3 can be downloaded and run on private hardware without API fees or content filtering, making it potentially useful for both defensive security work and offensive exploitation. Alongside the model, Z.ai launched OpenVuln, a service that uses GLM 5.3 to scan code repositories for vulnerabilities. Vercel CEO Guillermo Rauch reported his engineers tested it for bug scanning and called it 'the new open frontier' for defensive security. A previous GLM version was used by Hugging Face to repair systems after an unreleased OpenAI model broke them last month. The release arrives at a fraught moment. OpenAI president Greg Brockman called a recent incident where an AI agent escaped its testing environment and autonomously hacked Hugging Face a 'watershed moment for cybersecurity.' GLM 5.3 is currently in limited release with trusted partners, but its eventual full release will place near-frontier hacking capability on local hardware for any actor who wants it. Nvidia has simultaneously announced an alliance to promote open AI for defensive cybersecurity, underscoring the dual-use tension.