As enterprises give AI agents greater access to critical systems, security teams are looking beyond authentication to control what those agents can actually do. 

[For more news, click here]

Enterprise security has traditionally treated access as a point-in-time decision. A user, application or machine proves its identity, receives the permissions associated with that identity, and the organization monitors what happens from there. Autonomous AI agents are making that model harder to rely on as they gain the ability to interact directly with databases, cloud infrastructure and other business systems. An agent can have valid credentials and legitimate access to an environment while still taking an action outside the task it was assigned, creating a security problem that begins after authentication rather than before it. 

Delinea will showcase its runtime authorization capabilities for AI agents at GISEC Global 2026 in Dubai. The technology is designed to apply policy while an agent is operating inside a session, allowing organizations to permit, block or escalate an action before it is executed. It supports environments including databases, SSH hosts, Kubernetes clusters and cloud consoles.  

“AI adoption is moving the identity-security conversation beyond the moment of authentication,” said Mortada Ayad, VP Sales at Delinea. “The question for organisations is no longer simply whether an AI agent can access a critical system. It is whether they can govern what that agent does once it is inside. At GISEC, we want to put that challenge, and the need for runtime control, at the centre of the regional conversation.” 

As AI moves from systems that generate recommendations into workflows where software can carry out tasks on a user's behalf, the distinction becomes more practical. An agent may be authorized to connect to a database without every query it makes being appropriate for the task it was given. Runtime authorization adds a policy decision at the point of execution, allowing security teams to evaluate activity while the agent is working rather than relying entirely on permissions established when access was granted or reviewing the activity afterward. Delinea introduced the capabilities in July, combining just-in-time, task scoped access with real time policy enforcement and action level recording across the systems agents use. Tying permissions to the task being performed can also make it easier to contain an error or compromised session by giving security teams an opportunity to intervene when an agent moves outside its intended scope. 

A Partner Led Approach in the Region 

Regional partners will be central to the company's presence at the event. Delinea will focus on engaging existing partners, developing new relationships and building the capabilities needed to support customers adopting identity security for AI and other non-human identities. Its Partner Advantage Program provides a framework for resellers and solution providers, with plans to expand to managed service providers, professional services, advisory and technology partners. 

“For a capability this consequential, technology alone is not enough,” Ayad added. “Customers need partners who understand their environment, their risk and how to translate security policy into practical control. Our focus at GISEC is to equip and grow that ecosystem, so organisations can move forward with AI, knowing that authorised access does not become unchecked action.” 

Runtime policies still need to reflect which systems an agent can reach, which actions are necessary for a particular task and when additional approval should be required. Partners that understand a customer's infrastructure and security environment can help translate those requirements into controls that fit existing systems, particularly as AI agents begin working across different databases, infrastructure and cloud environments. 

Related Articles 

AI Agent Governance Is Falling Behind Enterprise Adoption, Optro Research Finds 

Cequence Security Launches Agent Personas to Automate AI Agent Governance Across the Enterprise 

Liferay AI Hub Gives Middle East Enterprises a Low-Code Path to Governed AI Agent Deployment