Salt Security Enhances Agentic Security Platform with AI Detection and Response Tools
Salt Security has recently made significant strides in bolstering its Agentic Security Platform by integrating advanced AI Detection and Response (AI-DR) capabilities. This enhancement is particularly noteworthy as it is designed to synergize attacks targeting large language models (LLMs) with subsequent activities occurring across multiple cloud platforms, tools, and application programming interfaces (APIs).
The newly integrated AI-DR capabilities provide organizations with real-time protection against a variety of threats that arise as AI systems function. These threats include direct and indirect prompt injection, jailbreak attempts, and unsafe model behavior. In today’s landscape, where AI technologies are becoming increasingly pervasive, the protection offered by Salt Security is especially vital for safeguarding sensitive information and ensuring the integrity of business operations.
Prompt injection has been identified as the most significant risk in the OWASP Top 10 for LLM Applications 2026. Salt Security emphasizes that the threat extends beyond merely manipulating the AI model itself; as AI agents become integrated with business tools and systems, a malicious prompt could result in unauthorized actions or the exposure of delicate information through the infrastructure linked to an AI agent. Thus, ensuring robust security at every level is critical.
This expansion of the platform follows a revealing study conducted by Salt Security, which disclosed that a staggering 92% of organizations are not equipped with the advanced security maturity necessary for protecting agentic environments. The survey, which gathered insights from over 300 security leaders, found that 66% of participants had witnessed an increase in API growth exceeding 50% in the previous year, driven largely by the rapid adoption of automation and AI. Alarmingly, nearly half of the organizations surveyed reported that they had to delay production releases due to concerns over API security, while 32% had experienced actual incidents related to API security.
Even though 79% of boards and executive teams have heightened their focus on AI security risks, only 18% of the respondents expressed extreme confidence in their capacity to detect attacks that involve generative AI. This statistic underscores a substantial gap in preparedness and highlights the pressing necessity for improved protective measures.
With the introduction of AI-DR capabilities, Salt Security aims to provide enhanced defenses against threats across the agentic pathway. These capabilities are seamlessly integrated into the existing Agentic Detection and Response platform, thereby extending runtime protection to the LLM layer. The company’s Agentic Security Graph meticulously maps the interrelations between AI agents, models, Model Context Protocol (MCP) servers, tools, and downstream APIs. This hierarchical mapping enables security teams to correlate an attack targeting a model with any subsequent activity coming from an MCP or API, all within a unified platform.
For illustration, if an attacker were to manipulate a billing agent to unveil information about a refund tool hosted on an MCP server, this could subsequently lead to attempts to exploit that API. This may occur either via the agent itself or through a direct attack on the API, with the potential to issue unauthorized refunds. Salt’s platform is designed to link these various stages of an attack, empowering defenders to measure the attack’s origin, identify targeted systems, and evaluate which business processes could be at risk.
Furthermore, Salt Security’s expansive Agentic Security Platform amalgamates Agentic Detection and Response functionalities with Agentic Security Posture Management. The latter component serves to aid organizations in recognizing their AI agents and pinpointing associated security vulnerabilities. Simultaneously, the detection and response capabilities remain vigilant, monitoring attacks across both agents and the systems with which they interact.
In a bid to centralize existing AI controls, Salt has crafted the platform to function alongside AI guardrails that are already established through various cloud services, endpoint products, Secure Access Service Edge (SASE) platforms, AI gateways, and managed AI services. Through the Agentic Security Graph, security teams can visualize these varied controls and the agents they safeguard in a single interface, facilitating the identification of discrepancies or areas lacking runtime protection.
In instances where current tools fall short of providing adequate coverage—particularly concerning bespoke AI agents operating in Kubernetes environments—organizations can readily implement Salt’s native AI-DR capabilities, all while maintaining their prevalent security products and gateways.
Roey Eliyahu, the co-founder and CEO of Salt Security, articulated the gravity of the protection being provided: "An attack on an AI model can become an attack on the systems that run the business." He elaborated that the native AI-DR offered by Salt safeguards interactions involving LLMs while linking activities at the model level to downstream tools and APIs. This comprehensive visibility enables security teams to grasp the entirety of an attack, discern what is at risk, and identify protection gaps while still leveraging the existing security measures in place.
As a result of these advancements, the native AI-DR capabilities are now available within Salt Agentic Detection and Response, further solidifying the platform’s status as a key player in the realm of AI security.