SentinelOne Unveils Singularity ADR to Secure AI Agents
SentinelOne introduces Singularity Agentic Detection and Response (ADR) to provide visibility, governance, and real-time threat prevention for AI agents operating within enterprise environments.
SentinelOne has launched Singularity Agentic Detection and Response (ADR), a new platform designed to address the growing security challenges posed by AI agents. This offering aims to discover AI agents, enforce usage policies, monitor their behavior, and detect and neutralize malicious or unintended actions in real time. By integrating with SentinelOne's existing behavioral prevention and detection capabilities, Singularity ADR provides a robust foundation for securing AI operations across endpoints, cloud workloads, and the broader enterprise.
The proliferation of AI agents, capable of holding credentials, chaining tool calls, and operating autonomously, presents significant risks. As highlighted by recent incidents, these risks can stem from sanctioned agents deviating from their intended functions or from malicious actors exploiting them. SentinelOne's Singularity ADR builds upon its existing security portfolio, which already protects endpoints, identities, and cloud workloads, by specifically focusing on the unique behaviors and actions of AI agents during runtime.
Singularity ADR enhances SentinelOne's AI Security offerings, complementing products like SentinelOne Prompt Security and Purple. It allows organizations to govern which AI agents are permitted to run and what resources they can access, while also blocking known malicious actions at the point of interaction. Any agent behavior that bypasses these policy controls is then subjected to the same advanced behavioral detection engines that SentinelOne employs for traditional threats, ensuring that rogue agents are identified and stopped.
The new platform extends SentinelOne's earlier introduction of Prompt AI Agent Security, which focused on discovering agents and managing their tool and data connections. Singularity ADR elevates agent security by combining this governance layer with comprehensive behavioral detection and response capabilities. This integration provides security teams with a unified view of agent activity, correlating AI-driven actions with traditional endpoint events. Analysts can now see the complete chain of events, from user requests to AI assistant actions and subsequent system activities, all within a single interface, eliminating the need for manual reconstruction.
Key features of Singularity ADR include comprehensive agent and management platform controller (MCP) inventory, providing visibility into all agents, tools, and permissions across various platforms. It offers multi-platform agent coverage, extending beyond MCP gateways to include agents in Salesforce Agentforce, Microsoft Copilot Studio, and OpenAI ChatGPT workspaces. "Agent Traces" monitor and log every agentic event, making them available for visualization, retention, and analysis in the Singularity Data Lake. The "Unified Storyline" feature integrates agentic event behavior with traditional EDR data into a single execution graph.
Furthermore, Singularity ADR delivers robust policy and execution governance, enabling granular control over agent usage, user access, and resource reach, with approval workflows for high-risk tools. It provides in-line prevention against common attacks like prompt injection, jailbreaks, and credential exfiltration. "Intent drift detection" identifies discrepancies between an agent's intended actions and its actual behavior, while "Agent containment" allows for the isolation of a single rogue agent without disrupting the entire workflow. Finally, "Agentic Investigation and governed response" automates investigation and response actions within predefined customer boundaries.
SentinelOne's Chief Product Officer, Chris Corde, emphasized that securing AI agents involves more than just stopping malicious actors; it also means addressing the unintended actions of sanctioned agents. He stated that Singularity ADR leverages SentinelOne's AI-based behavioral capabilities to stop these rogue actions using the same trusted platform customers already rely on. Tomer Weingarten, CEO of SentinelOne, added that AI security must encompass all aspects of AI, including the environments it runs on, the models, and the agents acting on its behalf. Singularity ADR brings runtime protection to these agents, enabling businesses to adopt AI with confidence.