Cisco 将用 Claude Mythos 对 AI 智能体做红队测试
Cisco plans to use Claude Mythos to red-team AI agents
Cisco 要拿 Anthropic 的 Claude Mythos 当攻击者，去打自家和客户的 AI 智能体，找提示注入这类漏洞，但只打授权过的目标。做智能体安全测试的团队可以参考这套授权红队流程，和自建测试工具相比省了不少搭建成本。
Cisco 宣布计划采用 Anthropic 的 Claude Mythos 来对其 AI 智能体产品进行红队测试，测试范围限定在已获授权评估的客户自有目标。这意味着 Cisco 会用 Claude Mythos 主动攻击自家及客户部署的智能体系统，在正式上线前发现提示注入、越权调用工具等安全隐患。
这类做法属于自动化安全评估的典型场景：让一个强模型扮演攻击者，对另一个系统发起对抗性测试，从而暴露人工审计难以覆盖的漏洞路径。Cisco 强调只针对获得授权的目标执行测试，避免越界操作带来的法律与合规风险。
随着企业智能体部署量增长，安全验证正在成为采购流程中的硬性环节。Cisco 选择 Anthropic 的模型而非自建工具，反映大型网络设备厂商开始依赖第三方前沿模型补齐自身安全测试能力，这类模式可能被更多企业安全团队效仿。
Cisco plans to use Claude Mythos to red-team AI agents
Cisco said testing would be limited to customer-owned targets authorized for assessment.
This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Claude, Anthropic, Cisco, 智能体.
Source attribution: curated from Tech in Asia (@Minh Le). Cross-source corroboration: 11 independent sources have covered this development (including Techsauce, OpenRouter, kimmonismus), confirming industry-wide signal strength.
