Radware Introduces Agent Trust Management for Secure AI Interactions

MAHWAH, N.J., Sept. 29, 2026 /CourierPR/ -- Radware, a global leader in AI and application security and delivery solutions for multi-cloud environments, has introduced a new solution called Radware Agent Trust Management. This innovation aims to help businesses securely participate in the emerging AI agent economy by ensuring that trusted AI agents can interact with applications without exposing sensitive data or business processes to risks.

As AI agent traffic grows, organizations face a dilemma: blocking agents can result in lost business, while allowing them unrestricted access can expose applications to new risks. Radware Agent Trust Management addresses this challenge by directly capturing an AI agent's declared intent, continuously assessing whether their actions remain consistent with that intent, and controlling what they are permitted to do. This solution is designed to move businesses beyond broad allow-or-block decisions and more securely capture the value of agent-driven interactions.

The solution is based on a comprehensive approach that includes understanding AI agent activity and intent, identifying agents across the ecosystem, continuously assessing whether agents can be trusted, and controlling what agents are permitted to do. Key capabilities of Radware Agent Trust Management include:

- Understanding AI Agent Activity and Intent: Prompt-aware, session-level visibility provides insight into an agent's identity, declared intent, actions, and navigation path, supporting security teams in investigating activity and making informed policy decisions. - Identifying Agents Across the Ecosystem: Multi-signal identification combines request headers, client-side behavioral signals, and direct agent responses to identify and classify both Web Bot Auth-compliant and non-compliant agents. Standards-compliant agents are cryptographically verified. - Continuously Assessing Trustworthiness: Radware captures declared intent through active conversation and uses AI-based behavioral analysis, prompt profiling, and anomaly detection to assess whether an agent's activity remains consistent with that intent. Trust assessments evolve as new signals emerge. - Controlling Agent Permissions: Permission-based governance is designed to enable organizations to control access to specific workflows, providing more granular options beyond broad allow-or-block decisions. For example, an organization can permit an agent to browse while restricting sensitive actions such as login or checkout.

Radware Agent Trust Management is now available as part of Radware Cloud Application Protection services, extending Radware’s bot management capabilities to address the distinct visibility, trust assessment, and governance requirements of AI agents. The solution is designed to assist organizations in supporting emerging use cases while maintaining greater visibility and control over how AI agents interact with their applications.

According to Radware’s H1 Global Threat Analysis Report, 77% of organizations are actively deploying or implementing AI agents, yet only 17.2% report full visibility into the AI agents operating within their environments. AI agents are increasingly interacting with applications across industries and performing tasks that can involve sensitive data, transactions, and critical business processes. In e-commerce and retail, agents can research products and complete transactions; in financial services, they can interact with applications involving sensitive information and financial activities; and in travel and hospitality, they can search, make bookings, and interact with inventory, pricing, and loyalty programs. Radware Agent Trust Management is designed to assist organizations in supporting these emerging use cases while maintaining greater visibility and control over how AI agents interact with their applications.

Identifying an AI agent alone does not establish whether it can be trusted. Even a verified agent can have malicious intent, deviate from its stated purpose, or attempt harmful actions. Radware Agent Trust Management is designed to go beyond relying solely on identity and observed behavior by also directly capturing an agent's declared intent through active conversation and continuously evaluating whether its actions remain consistent with that intent. This helps organizations distinguish legitimate agent interactions from potentially harmful activity.