Onyx Security CEO谈AI代理守护企业安全Onyx Security CEO on AI Agent Guardians for Enterprise
核心要点：随着自主代理能力与风险指数级增长，企业需要专门的AI系统来监督它们。Onyx Security CEO Maxim Bar Kogan是以色列初创公司的联合创始人，该公司专注于构建代理来监控其他代理。他解释了传统安全工具为何无法应对AI。企业无法依赖人工介入来处理百万倍规模的操作，现有的身份和端点工具也缺乏对代理“思考”过程的上下文理解。Onyx训练小型专业模型来判断何时需要更聪明的监督代理介入，从而平衡成本、延迟和可靠性。“我们被允许查看大量历史数据……但企业今天不愿意让Anthropic或OpenAI获取这些数据。”长期来看，这延伸到机制可解释性和控制先进AI，利用以色列深厚的安全人才储备。
The Takeaway: Enterprises need specialized AI systems to oversee autonomous agents as their capabilities and risks grow exponentially. Onyx Security CEO Maxim Bar Kogan, co-founder of the Israel-based startup building agents to watch other agents, explains why traditional security tools fall short for AI. Enterprises can't rely on human-in-the-loop for million-x scale actions, and existing identity/endpoint tools lack context on what agents are "thinking." Onyx trains small specialized models to flag when smarter oversight agents should intervene, balancing cost, latency, and reliability. "We're allowed to look at a lot of historical data... but enterprises today are not willing to have Anthropic or OpenAI give that historical data." Long-term, this extends to mechanistic interpretability and controlling advanced AI, leveraging Israel's deep security talent.
查看原文 →