Artemis 推出 Orion-1 安全模型，称其攻击调查能力超越通用模型
原标题：AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap
AI 摘要
美国初创公司 Artemis Security 推出专用安全模型 Orion-1，可自主调查并响应企业网络攻击，将来自认证服务、云环境、终端、SaaS 应用和网络设备的日志数据关联到单一攻击者，并给出置信度评分与处置建议。在自建 Decision-Grade Readiness 基准上，Orion-1 在攻击重建任务得分 80.8，高于 Claude Opus 5.5、GPT-6 Sol、Grok 4.7 和 Kimi K3 等通用模型（次优 58.3）。文章同时指出，面向攻击者的通用模型护栏正阻碍防御方分析攻击数据，厂商转而向认证防御者提供受限的专用模型。
正文节选
AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap Key Points - Artemis Security has launched Orion-1, an AI model that investigates and responds to cyberattacks on corporate networks on its own. - It connects scattered log data from networks, endpoints, and cloud services to trace attacks back to a single actor and kick off countermeasures. - On the company's own benchmark, Orion-1 outperforms general-purpose models like Claude Opus 5.5 and GPT-6