Radware Launches Security Layer for AI Agents

Radware has launched a security layer designed for a web where autonomous AI agents increasingly interact directly with applications.

What happened

The company introduced Agent Trust Management.

The product identifies AI agents, captures their declared intent, compares that intent with observed behaviour and applies controls over what the agent is allowed to do.

The aim is to give application owners a middle ground between blocking automated traffic entirely and allowing agents unrestricted access.

Why it matters

Agents are different from conventional bots.

They may shop, log in, fill forms or complete transactions on behalf of legitimate users.

That makes simple bot-blocking approaches less useful because some automated activity is desirable.

The security problem becomes one of identity, intent and permissions.

A business needs to know whether an agent is doing what the user authorised it to do.

The bigger picture

The agentic web will require new trust infrastructure.

Payments companies need to know who authorised a transaction, applications need to know what an agent is allowed to access and users need visibility into what happened on their behalf.

Radware's product reflects an emerging cybersecurity layer built specifically around machine actors.

As autonomous agents become more common, that layer could become as fundamental as bot management and identity security are today.