Anthropic report warns of GLM-5.3 cyber capabilities and safeguard risksMachine translation
Anthropic’s report says GLM-5.3 succeeded in 50 of 410 ExploitBench attempts, compared with 56 for Claude Mythos Preview. It also says researchers got the model to continue with attacks in 92% of simulated tests by prefilling its reasoning, and calls for independent safety testing.
报告同时给出漏洞利用测试结果和模拟测试中的防护绕过比例，为评估模型能力与滥用风险提供了具体依据。
Anthropic report warns of GLM-5.3 cyber capabilities and safeguard risks
Anthropic’s report says GLM-5.3 succeeded in 50 of 410 ExploitBench attempts, compared with 56 for Claude Mythos Preview. It also says researchers got the model to continue with attacks in 92% of simulated tests by prefilling its reasoning, and calls for independent safety testing.
报告同时给出漏洞利用测试结果和模拟测试中的防护绕过比例，为评估模型能力与滥用风险提供了具体依据。
材料 a5483d7311c44ff5ab418b1ccfe7b70c；建议 5f9fd04704644d2a9c7ef2649cec0f70；系统证据核验通过，非人工审稿。
Read at the original source