思科如何用OpenAI以机器速度防御
思科结合了网络智能和OpenAI模型，以构建更多的适应性防御。
Cisco建立了各组织用来连接和保护其系统的网络和安全基础设施。 随着对手利用AI以机械速度进行自动攻击和开发，捍卫企业基础设施必须不断发展。 让思科庞大的网络遥测与前沿AI推理对齐，OpenAI以零数据保留、严格的客户数据界限和可核查的隔离为中心，对企业数据隐私采取了严格的办法，这意味着思科可以自信地将前沿情报部署在对任务至关重要的系统之间，而不损害客户的优先事项或敏感的遥测。
Cisco的防御始于威胁移动的地方：网络。 Patel描述了从一个单一的安全周界向分散在微观服务、集装箱和集群之间的控制转变。 “我们可以在每个微型服务、每个Kubernetes容器和每个Kubernetes集群前放置一个小型安全堆，并将安全装置装入网络结构中。”
应用程序级别的控制可以限制通信到每个服务需要的连接，减少攻击者可以用来传播的路径。 安保小组还可以更有选择地隔离受影响的系统，使未受影响的服务在调查和应对时能够持续运行。
将网络情报与OpenAI的模型结合起来，Cisco就可以在机器尺度上将信号联系起来，Patel 称。 “孤立地说，两个信号可能看起来是良性的，但它们共同可以揭示一种有意义的模式。 网络情报使我们能够将整个网络的数据联系起来，并获得深入的见解，为安全行动提供信息。”
Patel指出维权者面临的一项紧迫挑战：攻击者能够利用脆弱性的速度与组织需要多长时间来弥补这一脆弱性之间的差距。 他认为大赦国际是帮助缩小这一差距的一种方式，使维权者有能力调查更多的威胁并作出更快的反应。 “AI手攻击速度和规模。 使用AI进行防御是保持安全的前提，"他称。
他将"日出"和"GPT‐6 Astra"等举措描述为加强思科及其客户安全的进步。 Patel 称指出
“日间休息和GPT-6 Astra不仅是我们的安全态势，而且是我们客户的安全态势的改进。”
帕特尔看到了一种通过的方法，它建立在各小组已经用来发现和解决脆弱性的过程之上。 “这些能力可以融入你现有的扫描工作流程
随着agents变得更加自主，安全小组还需要区分恶意干扰与agent以意想不到的方式追求合法目标。 需要安全遥测的Patel 称与可观察性数据并列，因此团队可以同时评估威胁和agent行为。
他概述了三个优先事项：保护agents免受恶意攻击，发现和拦截意外行为，并近实时运作。 这些优先事项正在引导人们转向能够随着agents行动而适应的政策和控制。
思科和OpenAI正通过日间防御网络探索如何帮助安保团队更快的分解曝光，并在思科云控制中执行正确的控制。 对帕特尔来说，这可以帮助将优势转移到维权者身上。
“OpenAI正通过允许我们以机器的速度和规模运作，将优势转移到捍卫者身上。”
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
How Cisco defends at machine speed with OpenAI
Cisco combines network intelligence and OpenAI models to build more adaptive defenses.
Cisco builds the networking and security infrastructure that organizations rely on to connect and protect their systems. As adversaries leverage AI to scale automated attacks and craft exploits at machine-speed, defending enterprise infrastructure must evolve. Pairing Cisco’s vast network telemetry with frontier AI reasoning gives defenders the toolkit necessary to outpace emerging threats. OpenAI’s rigorous approach to enterprise data privacy—centered on zero data retention, strict customer data boundaries, and verifiable isolation—means Cisco could confidently deploy frontier intelligence across mission-critical systems without compromising customer priorities or sensitive telemetry.
Cisco’s defense starts where threats move: the network. Patel describes a shift from a single security perimeter to controls distributed across microservices, containers, and clusters. “We can put a mini security stack in front of every microservice, every Kubernetes container, and every Kubernetes cluster, and fuse security into the fabric of the network,” he says.
Controls at the application level can restrict communication to the connections each service needs, reducing the paths an attacker can use to spread. Security teams can also isolate affected systems more selectively, allowing unaffected services to keep running while they investigate and respond.
Combining that network intelligence with OpenAI’s models allows Cisco to correlate signals at machine scale, Patel says. “Two signals might seem benign in isolation, but together they can reveal a meaningful pattern. Network intelligence allows us to correlate data across the network and derive insights to inform secure operations.”
Patel points to a pressing challenge for defenders: the gap between how quickly attackers can exploit a vulnerability and how long organizations take to patch it. He sees AI as a way to help narrow that gap, giving defenders the capacity to investigate more threats and respond faster. “AI hands attackers speed and scale. Using AI for defense is a prerequisite for staying safe,” he says.
He describes initiatives like Daybreak and GPT‑6 Astra as advances that strengthen security for both Cisco and its customers. “We’re able to find and patch vulnerabilities more efficiently, while shortening the time needed for investigation, response, and remediation,” Patel says.
“Daybreak and GPT-6 Astra have been step function improvements, not just for our security posture, but for our customers’ security posture.”
Patel sees a path to adoption that builds on the processes teams already use to find and address vulnerabilities. “These capabilities can fit into your existing scanning workflows, so you don’t have to fundamentally change how your team works.”
As agents become more autonomous, security teams also need to distinguish malicious interference from an agent pursuing a legitimate goal in an unintended way. Patel says that requires security telemetry alongside observability data, so teams can assess both threats and agent behavior.
He outlines three priorities: protect agents from malicious attacks, detect and intercept unintended behavior, and operate in near real time. These priorities are guiding the shift towards policies and controls that can adapt as agents act.
Through the Daybreak Defense Network, Cisco and OpenAI are exploring how to help security teams triage exposure faster and enforce the right controls in Cisco Cloud Control. For Patel, that can help shift the advantage back to defenders.
“OpenAI is shifting the advantage to defenders by allowing us to operate at machine speed and scale.”
Auto-fetched and translated by AI Atlas. All rights belong to the original author.