AI agent governance is the new resilience mandate – News

As enterprises accelerate their move toward autonomous workflows powered by artificial intelligence, the urgent need for robust AI agent governance has become a critical operational mandate.

An agent’s potential relies on the fact that they can perform higher-level tasks, including assuming identities and executing business processes. But if that is compromised, attackers have effectively hijacked business operations, according to Bipul Sinha (pictured), co-founder and chief executive officer of Rubrik Inc. 

“As folks have gone into the agentic work, what they’re realizing is that these agents can do 10x more damage in 1/10th of the time,” Sinha told theCUBE. “They’re running business processes. They are assuming your identity. In some ways they are super human.”

Sinha spoke with Savannah Peterson, during theCUBE’s coverage of the Rubrik “Resilience for Everything: Cloud, Identity, AI” interview series. TheCUBE, News Media’s livestreaming studio, explored the security implications of agentic AI and strategies for maintaining resilience. (* Disclosure below.)

Defining AI agent governance strategies

To address the risks of agents being hijacked by bad faith actors, Rubrik has introduced Rubrik Agent Cloud to help organizations audit, monitor and control their automated workforce. Companies must treat agents akin to new hires, requiring strict handbooks and oversight, Sinha explains.

“Rubrik Agent Cloud is all about understanding what agents are running in your company,” Sinha said. ”What are they actually doing? What are the guardrails around them? And if they misbehave, how do you undo those agents?”

Of course, the real agentic challenge lies in the scale of deployment. Without proper visibility, the speed at which these tools operate can turn a minor error into a major catastrophe. Information technology leaders should avoid trying to “boil the ocean”, Sinha notes. Instead, they should focus on automating three or four specific workflows to establish a baseline for safety before scaling.

“In production, you need to have accurate AI,” Sinha said. “You need to have cost control. Otherwise, if you’re doing a lot of inferences, your cost will run away.”

Here’s the complete video interview, part of News’s and theCUBE’s coverage of Rubrik “Resilience for Everything: Cloud, Identity, AI” interview series:

(* Disclosure: Rubrik sponsored this segment of theCUBE. Neither Rubrik nor other sponsors have editorial control over content on theCUBE or News.)

Photo: News