Forbes: Silverfort’s Launch Signals The Start Of Agentic AI Security Arms Race

Forbes covered the emerging field of agentic AI security that explains how enterprises are grappling with securing autonomous AI agents as they rapidly integrate into business workflows across departments from sales and support to finance and legal operations. 

The article highlights how AI agents, which act autonomously rather than following simple scripts, are creating new security challenges that traditional identity and access management (IAM) systems weren’t designed to handle. Unlike human employees, these agents operate at machine speed, make real-time decisions, and can escalate their own access permissions, creating what security experts call “agentic anarchy” if not properly managed. 

The article includes insights from NetSPI President and CEO, Aaron Shilts, on the modern IT landscape that requires more advanced approaches to secure. 

+ + + 

We’re coming perilously close to having to either rename the HR department to Human-AI Resources or to give the CTO full custody over tomorrow’s workforce. Either way, one thing is clear: the AI agents have arrived, and they’re already reshaping work as we know it. What began with Devin in early 2024 has now snowballed into Salesforce’s Agentforce, the rise of LangChain-based custom workflows, and enterprise-grade deployments like PwC’s AgentOS. Agentic AI, autonomous or semi-autonomous AI systems acting on behalf of a user, is rapidly becoming the tip of the spear of AI adoption, and one can only imagine how quaintly outdated our views from June 2025 will look within just a year’s time.

While the Agentic AI curve is rising fast, one question is threatening to drag it all down lying right beneath the surface: how do we manage, govern, and secure these agents at scale? 

If the last tech wave brought SaaS sprawl and death by a thousand point solutions, this one is threatening us with a future of agentic anarchy unless we play our cards right. 

As Aaron Shilts, CEO of NetSPI, puts it: “The attack surface has multiplied with the advent of Agentic AI and every AI agent with access to internal systems becomes a new entry point.”

“It’s like handing out admin credentials to enthusiastic interns who never sleep, don’t ask questions, and can spin up a thousand API calls before you even notice. That’s a red team’s dream,” Shilts continues. 

You can read the full story here.