AccuKnox Launches AI-Security 2.0 to Extend Zero Trust Protection to AI Models and Agents
嗯,用户让我用中文总结一下这篇文章的内容,控制在一百个字以内,而且不需要以“文章内容总结”或者“这篇文章”这样的开头,直接写描述即可。首先,我需要仔细阅读文章,抓住关键信息。 文章主要讲的是AccuKnox在2026年的RSA大会上发布了AI-Security 2.0平台。这个平台是一个基于身份的零信任框架,专门用于保护AI模型、代理和数据。它包含八个模块,其中六个已经可用,另外两个还在测试中。 接下来,我需要提取这些关键点:发布者、时间、产品名称、定位(零信任框架)、功能(保护AI模型和数据)、模块数量及状态。然后把这些信息浓缩到100字以内。 要注意不要使用任何开头语,直接描述内容。同时要确保语言简洁明了,涵盖所有重要信息。 可能的结构是:AccuKnox在RSA 2026发布AI-Security 2.0,作为基于身份的零信任框架,用于保护AI模型和数据。平台包含八个模块,其中六个已上线,两个测试中。 这样应该符合用户的要求了。 </think> AccuKnox在2026年RSA大会上推出AI-Security 2.0平台,作为基于身份的零信任框架,专为保护AI模型、代理和数据设计。该平台包含8个模块,其中6个已上线,另两个处于测试阶段,提供全面的风险管理和实时安全防护功能。 2026-3-23 18:8:23 Author: securityboulevard.com(查看原文) 阅读量:9 收藏

AccuKnox launched AI-Security 2.0 at RSA Conference 2026, positioning the platform as an identity-powered, Zero Trust framework built specifically for securing AI models, agents, and data. The release includes eight integrated modules, six of which are generally available and two in beta.

The GA modules cover the core risk surface organizations encounter when running AI in production. AI Security Posture Management (AI-SPM) continuously discovers and maps model endpoints, MLOps pipelines, and agent toolchains, giving security teams a real-time inventory of what is running, where, and who owns it. AI Detection and Response (AI-DR) analyzes CloudTrail, Azure logs, and runtime telemetry to surface risky activity, including unexpected resource creation, new region deployments, public exposure, privilege escalation, and abnormal GPU usage. AI Red Teaming automates adversarial testing against hallucinations, prompt injection, toxicity, and unsafe code generation. Prompt Firewall and AI Pen Testing round out the GA lineup, while AI-DSPM and AI-GRC remain in beta, along with AI Identity Controls.

The platform supports upstream caller sequence-aware authorization for multi-agent workflows and offers runtime sandboxing with least-privilege tool execution. It uses open standards including SPIFFE and OpenFGA and integrates with Amazon Bedrock, Gemini, Ollama, and vLLM. Deployment spans AWS, Azure, GCP, on-premises Kubernetes, VMs, containers, and serverless environments, and AccuKnox is available on the AWS, Azure, Red Hat, and Oracle Cloud Marketplaces.

“AI is quickly becoming the core infrastructure for modern enterprises, but the security model around it is still catching up. AccuKnox AI-Security 2.0 takes a meaningful step forward by applying Zero Trust principles directly to the AI layer, bringing visibility, runtime protection, and identity-driven governance to the environments where AI actually runs,” said Golan Ben-Oni, CIO and CISO of IDT Corporation.

AccuKnox was incubated at SRI International and is backed by National Grid Partners, Dolby Family Ventures, Dreamit Ventures, Avanta Ventures, and the 5G Open Innovation Lab. The platform builds on CNCF open-source projects KubeArmor and ModelArmor.


文章来源: https://securityboulevard.com/2026/03/accuknox-launches-ai-security-2-0-to-extend-zero-trust-protection-to-ai-models-and-agents/
如有侵权请联系:admin#unsafe.sh