News
AI Agents Are Creating a New Job for the Enterprise Firewall
Enterprise firewalls have spent decades deciding which network traffic to allow through. The rise of AI agents is giving them another job: figuring out what those agents are doing and which systems they should be allowed to reach.
Gartner's new September 7 Magic Quadrant for Hybrid Mesh Firewall report shows AI security becoming a growing part of the enterprise firewall market, as vendors add controls designed specifically for AI applications, agents and the infrastructure running them.
The research firm identifies three emerging areas where hybrid mesh firewall vendors are developing capabilities AI usage control, AI runtime security and AI workload security.
Those capabilities include controlling access to AI applications, defending against prompt injection and protecting the containers, virtual machines and clusters supporting AI workloads.
The report evaluates 12 hybrid mesh firewall vendors. Gartner defines the category as cloud-managed firewall platforms that provide security across physical, virtual, and cloud environments. But its definition also shows how quickly the role of those platforms is expanding
Among the optional capabilities Gartner now identifies for hybrid mesh firewalls are discovering AI application usage, logging that activity, preventing certain data transfers, and assessing risk at the network level.
The research firm also points to the ability to discover and control third-party AI application traffic through APIs, including Model Context Protocol (MCP) and agent-to-agent (A2A) connections.
That matters as enterprise AI moves beyond employees interacting with chatbots.
Agentic systems can connect AI models with applications and services, creating machine-to-machine traffic that security teams also need to govern
The shift is already visible in vendors' product plans.
Gartner said Check Point is integrating technology from its Lakera AI acquisition into its container, cloud, and on-premises firewall capabilities, providing protection for prompts as well as MCP and A2A communications.
The company is also focusing on protecting AI data centers and GPU clusters.
Cisco, meanwhile, is focusing on protection for AI agents and applications. The researchers highlighted Cisco AI Defense's AI Runtime Protection, which is designed to automatically configure guardrails against attacks including prompt injection, model extraction, and denial of service.
Fortinet plans to introduce a unified MCP framework for AI agent-based automation over the next 12 months, alongside workflow automation using natural-language interfaces.
Palo Alto Networks is also pushing deeper into agentic AI security. Gartner said the vendor plans to focus on "intent-driven autonomous AI agent security," agentic endpoint protection, and protection for vibe coding.
Its existing Prisma AIRS offering provides AI runtime security. AI isn't only becoming something firewalls need to protect. It's also increasingly being built into the tools administrators use to operate them.
All 12 vendors evaluated by Gartner offer LLM chat assistants, although their maturity varies considerably.
Vendors are using natural-language interfaces for tasks including policy management, threat prevention and asset discovery, while adding automated troubleshooting, predictive monitoring, conflict detection and policy analysis.
Actual enterprise adoption of those capabilities appears to be moving more slowly.
The research and advisory company said customers are not yet fully using AI-based orchestration and LLM assistants for everyday firewall administration, limiting their impact on policy optimization.
Still, the firm sees the market moving toward more automated and "intelligence-driven" security platforms.
It points to growing use of AI for threat detection, predictive analysis, and policy recommendations alongside deeper integration with XDR, NDR, and EDR systems.
The result is an expanding definition of the enterprise firewall. Protecting network traffic remains its central job.
However, Gartner's research suggests the next generation will increasingly be expected to understand AI traffic as well - - including which AI applications are being used, what data is moving through them and how autonomous agents communicate with other services.