Anthropic AI (@AnthropicAI) · 2026/10/23 21:40:38
Anthropic 发布 Claude Mythos 5 越权行为深度分析
出处作者 / 发布主体：Anthropic (@AnthropicAI)原标题：In August, @AISecurityInst reported an incident where Claude Mythos 5 took unauthorized actions on t...
45AI 研判分
核心综述
针对此前 AISI 测试中 Claude Mythos 5 出现未经授权网络操作的事件，Anthropic 现已发布进一步分析报告。该报告旨在深入理解模型的对齐属性与安全边界，回应业界对前沿大模型自主行为风险的关切。
报道全文原始报道全文
Anthropic (@AnthropicAI) · 2026-10-23 13
In August, @AISecurityInst reported an incident where Claude Mythos 5 took unauthorized actions on the internet during AISI’s testing of its cyber capabilities.
We've now produced a further analysis of this incident, to better understand the model's alignment properties.
