Unveiling the Vulnerabilities: How AI is Exploiting MCP Servers as a Prime Target

AI adoption is rapidly increasing, leading to the evolution of the infrastructure that supports it. With new standards and connectors emerging at a fast pace, the ecosystem is constantly evolving. This rapid evolution enhances the capabilities of AI but poses challenges for security teams in maintaining adequate protections.

One notable example of this is the rise of MCP servers as a preferred standard for connecting AI agents to external tools and data within a short span of 12 months. By December 2025, MCP servers were widely adopted by major coding assistants and leading LLMs. Despite its rapid growth, the adoption of MCP has outpaced the development of security solutions, leaving vulnerabilities that need to be addressed.

To address the security gaps, cybersecurity vendors are developing products such as “AI firewalls” specifically designed to protect AI models, agents, and the tools they connect to from threats like prompt injection and data leakage. Check Point’s AI Network Firewall, released in July 2026, falls into this category, focusing on defending AI infrastructure from emerging threats.

The need for AI firewalls is particularly evident in the case of MCP servers, which serve as connectors for AI agents to access external tools and data. The rapid growth of MCP servers has been remarkable, with Anthropic launching MCP as an open standard in November 2024 and quickly gaining widespread adoption by leading AI platforms and coding assistants by December 2025.

However, along with its advantages, MCP servers introduce new attack surfaces that traditional AI defenses are not equipped to handle. Threats such as tool poisoning, rug pull attacks, and data exfiltration have been identified, highlighting the importance of implementing robust security measures for MCP servers.

While securing MCP servers is crucial, it is just one aspect of safeguarding AI infrastructure. Security measures need to extend beyond MCP servers to protect AI agents from various threats. Companies like TrueFoundry and Cisco offer security solutions that address the unique challenges posed by MCP servers within the broader context of AI security.

Check Point’s AI Network Firewall takes a network-centric approach to integrating AI security into existing firewall infrastructure, providing protection for AI applications, agents, and interactions with MCP servers. As the AI infrastructure continues to scale rapidly, it is essential for security teams to stay ahead of emerging threats and implement comprehensive security solutions to safeguard AI ecosystems effectively.