Check Point launches AI Network Firewall to secure prompts, agents, and AI enterprise activities

Check Point introduces industry-first AI Network Firewall

Check Point Software Technologies has launched the AI Network Firewall as part of firewall software release R82.20, marking the first time AI security capabilities have been embedded directly into physical firewalls that organizations already operate 1

1

2

2

. This AI-specific security solution addresses a fundamental problem: traditional firewalls were never designed to inspect or govern the new class of network traffic generated by AI systems, including prompts, autonomous agent actions, and model interactions 3

3

. The solution operates through Check Point's AI Defense Plane, requiring no new infrastructure or network redesign 4

4

.

Closing the AI blind spot in enterprise networks

The urgency behind this launch stems from alarming findings in the AI Security Report 2026 published by Check Point Research. Between 87% and 93% of organizations experience at least one high-risk generative AI interaction every month 2

2

. Even more concerning, the share of prompts carrying sensitive corporate, personal, or regulated data doubled in just one year to one in every 25 interactions 3

3

. Organizations now run an average of ten AI applications per month, many operating outside any formal governance process as shadow IT 2

2

. This visibility gap represents a growing threat as employees send prompts to generative AI platforms, applications make model calls behind the scenes, and autonomous agents communicate with systems without human involvement 1

1

.

Three-domain protection for AI enterprise activities

The AI Network Firewall delivers protection across three critical domains. For employee AI use, it discovers both sanctioned and shadow AI apps, provides visibility into prompt use cases and intents, governs access to approved tools, and prevents sensitive data from leaving the network based on context 4

4

. The Model Context Protocol (MCP) protection discovers MCP communication, provides full visibility into servers and tools, and enforces access policies across every interaction—a crucial capability given that Check Point Research found security weaknesses in 40% of 10,000 MCP servers reviewed 2

2

. For AI applications and large language models, the firewall prevents prompt injection attacks and adversarial inputs by blocking malicious prompts inline before they reach the LLM, with no application changes required 4

4

. Check Point Research identified 15,300 indirect-injection payloads planted in public web pages, with roughly 70% hidden in parts of the page users never see 2

2

.

Network control plane becomes central to AI security

The network has emerged as the central hub where all AI activity converges, making it the logical control point for real-time AI security enforcement 1

1

. "AI is transforming the enterprise network, and with the AI Network Firewall, we are transforming the firewall to secure it," said Nataly Kremer, Chief Product Officer at Check Point 3

3

. By understanding prompts, model interactions, file uploads, API calls, and agent behavior in real time, the solution provides visibility and control that organizations need to safely adopt AI at scale 1

1

. This approach prevents data exfiltration, detects API abuse, governs MCP servers, and maintains centralized oversight across employees, applications, and autonomous agents 1

1

.

Integration with AI Defense Plane and multi-cloud environments

Unlike alternatives requiring separate virtual firewalls deployed alongside existing infrastructure, Check Point delivers protection directly from physical or virtual firewalls customers already operate, scaling across branches, data centers, cloud, and multi-cloud environments 2

2

. The AI Network Firewall integrates into Check Point's AI Defense Plane, a unified control plane for discovering, governing, and protecting AI across networks, endpoints, cloud, applications, and APIs 4

4

. Additional enforcement points span standalone API for self-managed applications, endpoint protection for employees, containerized firewalls for AI data centers, and WAF capabilities 4

4

. Check Point is also extending central policy management to SASE and SD-WAN, enabling zero-trust enforcement across IT, OT, and micro-segmentation tools from a single console 4

4

. Pete Finalle, Research Manager at IDC, noted that "the native integration of AI security across existing enforcement points provides improvements to visibility, telemetry effectiveness, security posture, and management simplicity" 2

2

. The solution supports human-language policy management and automated event analysis, enabling security teams to operate at AI speed while reducing operational complexity and human error 1

1

.