Salt Security Adds AI Detection, Response to Agentic Security Platform
The company says its new runtime monitoring maps suspicious LLM activity across connected tools, MCP servers and APIs.
- By Danielle Naidu
- Sep 24, 2026
With its latest platform update, Salt Security has introduced native AI Detection and Response for monitoring large language models during operation. The capabilities are available through Salt Agentic Detection and Response (AG-DR).
Salt said the new monitoring looks for jailbreak efforts and model activity it flags as unsafe. It also covers direct and indirect forms of prompt injection, along with additional threats that emerge during live use. Salt connected the capabilities to its Agentic Security Graph, which maps activity involving models, agents, tools, Model Context Protocol servers and APIs.
The OWASP Top 10 for LLM Applications 2026 places prompt injection at the top of its risk list. Salt also cited its 2H 2026 State of Agentic AI and API Security survey, which found that 49.8% of organizations had confirmed or suspected an AI agent performed an unintended or unauthorized action during the previous year. Just 12.5% said they had dependable, end-to-end visibility into an agent’s activity, beginning with the prompt and continuing across every MCP server and API involved.
The update also consolidates information from AI guardrails and gateways already used by customers. Salt said Agentic Security Posture Management (AG-SPM) uses that information to display agent coverage and locate security gaps. AG-SPM is used to examine agents and identify security gaps; AG-DR provides runtime attack monitoring across agents and connected systems, including internally developed agents operating on Kubernetes.
About the Author
Danielle Naidu is assistant editor for Security Today, Campus Security Today, Occupational Health & Safety and Environmental Protection Online.