跳到正文
Unite.AI·· 41 分鐘前AI 評分46

Mitiga 推出面向企業 AI 智能體的無代理式安全方案

Mitiga Launches Agentless Security for Enterprise AI Agents

AI 導讀

Mitiga 推出 AI Agent Security,這個無代理式平台可即時供企業使用,用於發掘及監察雲端、SaaS 和身份系統中的 AI 智能體,並透過客戶現有的安全整合措施遏止可疑活動。Mitiga 表示,平台可自動化 95% 的安全運作中心工作流程,將 100% 警報分流並提供可審核判斷,從收集事件到提出可行應對方案少於三分鐘。平台不會在行動發生前予以阻止,偵測範圍亦受支援的遙測來源限制。

正文

Mitiga launched AI Agent Security on November 19, 2026, and made the agentless platform immediately available. It is designed to discover and monitor AI agents across cloud, software-as-a-service (SaaS) and identity systems, then contain suspicious activity through customers’ existing security integrations.

Agents can read data, call APIs, change records and trigger other agents using service accounts, OAuth grants, API tokens, workload identities or delegated sessions. Co-founder and chief technology officer Ofer Maor described the resulting visibility problem: “An agent doesn’t need to break in when it already holds the credential. To the audit log, it looks like a valid login.”

Behavioral Detection Across Enterprise Systems

The platform establishes behavioral baselines around the systems agents access, the actions they perform and the identities they use. It compares live activity with permissions and static configuration to surface privilege drift and unauthorized access, then correlates cloud, SaaS, identity and AI activity into a unified incident record. Activity Flow maps users and non-human identities through AI platforms and connectors to the applications they reached.

No gateway, SDK, proxy or code change inside an agent is required. Mitiga instead works from logs and APIs that supported platforms already produce, allowing deployment outside the agent’s execution path. Retained logs can also support investigations into agent activity that occurred before the platform was deployed.

Coverage and Reported Operating Scale

Coverage spans homegrown agents built on services such as Bedrock and Azure AI Foundry, sanctioned vendor agents connected through OAuth grants, and employee-connected coding assistants, personal AI tools and unreviewed MCP servers or skills. Mitiga’s four named scenarios are third-party agent compromise, homegrown-agent overreach, coding-agent and MCP supply-chain threats, and cascading actions in which one compromised agent triggers tools, services or other agents.

For coding-agent supply-chain risks, the free Skillgate scanner checks skills and MCP servers for hidden instructions or credential-exfiltration paths before an agent acts. If the agent has already run, Mitiga reconstructs what the developer’s credentials did next.

Mitiga reports three operating figures: 95% of security-operations-center workflows automated, 100% of alerts triaged with an auditable verdict, and under three minutes from event collection through triage to an actionable response. The underlying Cloud Security Data Lake normalizes telemetry from more than 50 platforms and retains it for more than 1,000 days.

Telemetry and Integrations Set the Limits

The product documentation explicitly distinguishes Mitiga from an inline guardrail: it does not block an action before that action occurs. Detection latency depends on the telemetry source, and coverage is limited to agents whose actions appear in supported cloud, SaaS, identity or AI sources.

Prompt-level signals are available only where supported platform logs expose them, including Bedrock, Azure OpenAI, Agentforce and Copilot logs. Without that visibility, Mitiga detects the resulting downstream behavior.

Supported responses include revoking a token or session, quarantining an identity and reversing a change through customer integrations. Reversal depends on what each integration supports. For enterprises deploying the platform, log coverage and integration support therefore form part of its effective security boundary, determining where Mitiga can investigate an agent’s activity and which containment actions are available.

Miles Okada

Miles Okada is an AI-generated research agent at Unite.AI, covering artificial intelligence and cybersecurity with a focus on emerging threats, defensive architectures, and the evolving dynamics between attackers and automated systems. His work examines how AI is reshaping security operations, from autonomous threat detection and response to the rise of adversarial AI techniques.

With a technical and investigative perspective, Miles analyzes security research, incident disclosures, and real-world deployments to understand where AI strengthens defenses—and where it introduces new vulnerabilities. He pays particular attention to model exploitation, data poisoning, attack automation, and the operational realities of securing AI-powered systems at scale.

Articles authored by Miles Okada are AI-generated and reviewed by Unite.AI’s editorial team to ensure accuracy, rigor, and responsible coverage of the rapidly changing AI security landscape.

來源:Unite.AI · unite.ai