NEWS · MODELS · #1267
Anthropic says Zhipu's GLM-5.3 nearly matches Claude Mythos Preview for exploit development
Anthropic's Frontier Red Team reports that Zhipu AI's open-weight model GLM-5.3 can autonomously build cyber exploits close to the capability of Claude Mythos Preview and that GLM-5.3 shipped without effective safeguards. On ExploitBench GLM-5.3 produced working Chrome V8 exploits in 50 of 410 attempts versus Mythos's 56, Anthropic showed the model chained novel vulnerabilities with limited human help, and an independent CAISI assessment reached similar conclusions.
KEY POINTS
- Anthropic's Frontier Red Team reports that Zhipu AI's open-weight model GLM-5.3 can autonomously build cyber exploits close to the capability of Claude Mythos Preview and that GLM-5.3 shipped without effective safeguards.
- On ExploitBench GLM-5.3 produced working Chrome V8 exploits in 50 of 410 attempts versus Mythos's 56, Anthropic showed the model chained novel vulnerabilities with limited human help, and an independent CAISI assessment reached similar conclusions.
- If accurate, the report shows open-weight frontier models can materially enable offensive cyber capabilities, affecting access controls, defensive tooling, and policy.
WHY IT MATTERS
If accurate, the report shows open-weight frontier models can materially enable offensive cyber capabilities, affecting access controls, defensive tooling, and policy.