Delinea Launches Runtime Authorization to Secure Unpredictable AI Agent Actions

Agent security in focus for Delinea

Delinea targets AI agent risks with real-time authorization As AI agents gain access to sensitive enterprise systems, AI agent security is becoming increasingly difficult to manage with traditional identity and privilege controls. The shift is driving demand for real-time authorization and stronger auditability without slowing productivity. Traditional identity and governance models were built around persistent human users and narrowly defined machine identities. AI agents create a new challenge because they can rapidly spawn temporary identities, access multiple systems and disappear once a task is complete, according to Art Gilliland (pictured), chief executive officer of Delinea Inc. "The typical gain visibility, scan for everything, set policy, and then enforce, that just is not going to work in AI," Gilliland said. "In fact, I think you're wasting your time looking for and trying to discover all the agents. And so you just need a different approach." Gilliland spoke with Krista Case at Black Hat USA, during an exclusive broadcast on theCUBE, SiliconANGLE Media's livestreaming studio. They discussed strengthening AI agent security through real-time authorization and eliminating standing privileges without disrupting enterprise workflows. (* Disclosure below.) AI agent security needs new response Trying to identify every AI agent, create accounts for them and issue credentials would be impractical. Enterprises need a different approach that can secure agent activity without slowing operations. "The approach that we are taking, and we announced some technology last week about it and some customers that are using it, is to focus on the assets you care about," Gilliland said. "What are these agents going to touch? Where is the sensitive data in your environment? Where are the sensitive things in your environment?" Organizations need to distinguish between a human accessing a system and an AI agent acting with that person's credentials. They can then apply different permissions and evaluate each agent action in real time to determine whether it should be allowed. "That level of visibility, you can record it, you can watch it, so you have a way to demonstrate for compliance what the agent actually did," Gilliland said. "It's a very different approach." Real-time authorization sits in line within the protocol, giving organizations a full audit trail of agent behavior. Unlike static access controls, it can surface patterns across sessions -- making it useful not just for security, but for compliance and business operations as well. "You can decide. So this is just, in our opinion, a way better and more effective approach," Gilliland said. "So that you can go fast and not break stuff." Authorizing access at the front door is not enough, because the real risk comes from what the AI does next. It may scan through files in systems such as SharePoint, find additional credentials and use them to elevate its privileges, Gilliland noted. "That's the stuff that humans don't do because they're not going to go through every file, machines don't do because they're very discreet, but AI will do whatever it takes to solve the problem you ask it to solve," Gilliland said. "That is that sort of non-deterministic behavior with a machine sort of capability is why you need that sort of real-time authorization and not just front-door access." Here's the complete video interview, part of SiliconANGLE's and theCUBE's coverage of Black Hat USA: (* Disclosure: Delinea Inc. sponsored this segment of theCUBE. Neither Delinea nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Delinea Introduces Runtime Authorization to Secure AI Agent Actions

"The security industry spent years solving credential theft, but AI agents have introduced a new problem: authorized access doing unauthorized things," said Art Gilliland, CEO at Delinea. "You can have perfect credential hygiene and still have an agent tear through your production environment in milliseconds. Recording what happened or revoking access after the fact doesn't stop that - enforcing policy on the action as it runs does. The perimeter has moved to inside the session, and no one has figured out how to address that until now." Consistent policy and a complete audit record, regardless of how agents connect. Agents reach databases, SSH hosts, Kubernetes clusters, and cloud consoles directly, not only through an MCP server. The Delinea Platform authorizes and records each of those connections through a single control point, across every protocol agents use. Security teams get the same visibility and enforcement regardless of how the agent was deployed or which systems it reached. No standing credential for an attacker to exploit. The agent never holds a credential. The Delinea Platform injects it just-in-time at the moment of access, scoped to the task rather than inherited from the person who deployed it, and revokes it automatically when the task completes. Because the credential is never exposed to the agent, there's no lingering exposure between sessions. Policy enforced on every action, not just when the session opens. A single AI session can carry dozens of tool calls, each with a different risk profile. The Delinea Platform evaluates and enforces policy on each tool call individually - allowing it, blocking it, or pulling in a human before it executes. When an agent misbehaves, the blast radius stays contained. Policy built specifically for agents, not retrofitted from humans. The Delinea Platform identifies whether a connection is agent-driven or human-driven before granting access, applying agent-specific policy at the moment of connection so the right controls are in place before the first action runs. A complete, defensible record of every action, tied to a named identity. Every tool call, database query, and SSH command is recorded and accountable to a specific identity. When something goes wrong, teams have the control to investigate, respond, and stand behind every access decision that was made.