Silverfort unveils AI Agent Security to protect agentic identities, securing MCP deployments with inline, dynamic security controls

Identity and access security layer gives enterprises the confidence to rapidly and securely adopt AI agents and accelerate innovation

BOSTON, June 18, 2025 /PRNewswire/ -- Today Silverfort, the leading identity security company, released its latest innovation, AI Agent Security. The new product empowers CISOs and their organizations to securely adopt AI by controlling what AI agents are allowed to access and protecting their identities. It keeps them governed, visible, and protected with the same rigor applied to human users. Trusted by more than 1,000 organizations, including many of the world's largest enterprises, Silverfort has an advantage in solving this emerging need at enterprise scale. With Silverfort AI Agent Security, teams can tether every AI agent to a human identity, analyze its activity and the resources that it's trying to access, and dynamically enforce access control policies to prevent misuse and data leakage.

AI agents operate in a grey area between human and non-human identities, requiring purpose-built security to match their unique risks. AI agents act autonomously, make decisions, and access sensitive systems, frequently using delegated or privileged identities, which obscures accountability. Many organizations adopt the MCP protocol to allow AI agents to access corporate resources, but do so without a clear security framework, exposing critical systems to potential misuse. MCP enables highly dynamic, implicit access to internal resources based on model behavior. Without strict boundaries or observability, this can lead to unrestricted and unmonitored access across critical systems by AI agents that often act unpredictably.

"The pressure to adopt AI is growing rapidly. CISOs and CIOs face a task that feels impossible—aggressively adopt AI, yet be sure to keep the company's systems safe," said Yaron Kassner, CTO and co-founder, Silverfort. "Today AI agents connect straight from the LLM into corporate data stores with limited visibility or guardrails. By treating AI agents as a new type of identity, and connecting them to their human owners, we create an audit trail of activity and accountability. Further, security teams can apply inline security controls they need to allow teams to innovate and accelerate AI adoption, without the fear of destruction to corporate data and reputation."

Silverfort approaches securing AI agents the same way it secures human identities and service accounts: inline and in real time. Just as Silverfort protects authentication protocols like NTLM and OpenID Connect, it now protects MCP, too. Silverfort's model ties every agent action to a real human owner, enforces least privilege, and logs an immutable audit trail. Sitting inline between the AI agent and the MCP server, Silverfort inspects every call before it reaches the target resource, protecting all corporate systems with robust, real-time security controls and preventing AI agents from going beyond their intended purpose and causing damage.

Benefits include:

• Inline security controls: Dynamically grant or deny access of AI agents to MCP servers and other corporate systems, restricting over-privileged access. 
• Control of AI tools: Approve how and when MCP servers & API tools are used.
• Accountability and auditability: Link every AI agent action to the initiating human for clear accountability and to create an audit trail.
• Stop AI agent overreach: Prevent misuse and data leakage, whether malicious or accidental. Make it impossible for AI agents to act outside of their original purpose, and for attackers to use them for lateral movement or privilege escalation.

Combined with precise least-privilege access enforcement, AI agents are strictly limited to actions they are explicitly authorized to perform, so every decision aligns with both security policies and contextual relevance. Organizations can discover and classify AI agents, monitor their activity, and tie each one to a human owner. Designed for ease and quick deployments, Silverfort integrates with no impact to end users or the developers building the agents.

Learn more about Silverfort AI Agent Security and see how to safely innovate with AI agents, without compromise.

About Silverfort

Fueled by a belief that identity professionals deserve better, we found a way to break down the silos of identity security—eliminating the gaps and blind spots left behind by a patchwork of point solutions. The Silverfort Identity Security Platform is the first to deliver end-to-end identity security, protecting every identity in the cloud, on-prem, humans, machines, and everything in between. Our patented technology—Runtime Access Protection (RAP)—natively integrates with the entire IAM infrastructure, giving businesses visibility into all identities, analyzing every access, and extending active protection to resources that could not be protected previously—including NHIs, legacy systems, command line tools, and IT/OT infrastructure. It is easy to deploy and use, and doesn't disrupt business operations, resulting in better security outcomes with less work. Silverfort is the identity security platform that both identity and security professionals deserve, earning the trust of more than 1,000 leading organizations, including several Fortune 50 companies. Learn more at silverfort.com.

SOURCE Silverfort Ltd