Open-weight & owned stack · Benchmarks evals
Anthropic Reports GLM-5.3 Can Autonomously Build Cyber Exploits
Anthropic reported that Z.ai’s open-weight GLM-5.3 chained browser vulnerabilities and performed cyber tasks near the capability of restricted frontier models while often bypassing safeguards.
Read the original at anthropic.comOpens the publisher's site in a new tabAlso covering this
Quoting Anthropic Frontier Red Teamsimonwillison.net, Sep 29Anthropic raises alarm over elite hacking ability of Chinese firm Z.ai’s GLM-5.3scmp.com, Sep 29GLM-5.3 and the spread of advanced cyber capabilities \ Anthropicanthropic.com, Sep 29Anthropic Warns GLM-5.3 Builds Exploits Like Mythos | Trending Topicstrendingtopics.eu, Sep 30Alerta por débil seguridad del modelo chino GLM-5.3 de Zhipu AI - KCH FMkchcomunicacion.com, Sep 30Anthropic warns a free Chinese AI model can already build working hacks - Startup Fortunestartupfortune.com, Sep 29Anthropic says Zhipu's open-weight GLM-5.3 nearly matches Claude Mythos Preview at building exploitsthe-decoder.com, Sep 30Anthropic tests GLM-5.3 — its safeguards fall to… | AI/TLDRai-tldr.dev, Sep 29