Did CyberArk become Idira?

Yes. On May 12, 2026, Palo Alto Networks rebranded CyberArk as Idira and positioned it as the next-generation identity security platform for the AI enterprise. The underlying product portfolio is unchanged; the rebrand reflects Palo Alto Networks' platform consolidation strategy.

How is Delinea different from CyberArk?

Delinea delivers centralized authorization with runtime enforcement across human, machine, and AI agent identities in one platform, including a broker for modern workload access. CyberArk is positioned as a vault-centric platform with additional modules for cloud and agentic use cases. Delinea is designed to integrate with existing environments, while CyberArk is part of a broader Palo Alto Networks platform bundle.

Is Delinea cloud-native?

Yes. The Delinea Platform is cloud-native, with a 99.995% uptime SLA, 30 minutes from code to production, and a 2-minute mean time to recovery. Delinea also supports on-premises and hybrid deployments for regulated workloads.

Does Delinea support AI agents and non-human identities?

Yes. Delinea provides runtime authorization for AI agents through StrongDM, with MCP-native connectivity plus session and command auditing. Customers are using these capabilities in production, and AI capabilities are included in the Delinea Platform.

What does Delinea integrate with?

Delinea integrates with major identity providers such as Okta, Microsoft Entra, and Google Cloud Identity; cloud platforms such as AWS, Azure, and Google Cloud; Kubernetes; databases including Postgres, Redis, and MongoDB; DevOps pipelines; and security tools including SIEM, SOAR, and ITSM platforms.

How long does a Delinea deployment take?

Modern workload access deployments typically reach production in weeks. Traditional PAM deployments typically take less than 90 days. Delinea is positioned as requiring fewer resources to manage and less time to achieve full functionality than CyberArk.

How Agentic AI Security is Different and Why it Matters

Why agentic AI security matters

Enterprises in nearly every industry are deploying AI agents to automate operations and improve productivity. Gartner® predicts that "by 2030, 50% of enterprise application software offerings will include some agentic AI features, up from less than 5% in 2025".¹

Traditional enterprise identity security solutions were built to control human users, not agentic AI systems operating at machine speed and scale. While a person might access a handful of systems during the course of a day, an AI agent can interact with dozens or even hundreds of systems simultaneously, calling other agents and triggering actions across systems in seconds.

The risks of ungoverned AI agents

Many organizations are deploying AI agents faster and more widely than existing governance processes can handle. Today, most enterprises don't know how many AI agents are running in their environment, where those agents are deployed, or what systems they can reach. They have no way of knowing which AI agents pose the greatest risk. And once agents have access to an application or system, there is no way to control what actions they can perform.

In a 2026 IBM Institute for Business Value report, 77% of CIOs and CTOs said AI adoption is outpacing their current governance capabilities. And in Delinea's 2026 identity security report, 46% of respondents admit their identity governance is deficient around AI systems. When AI agents operate outside traditional identity controls, the attack surface expands in ways that are difficult to detect and even harder to contain.

The threat landscape is shifting as well. New offensive AI platforms such as CyberStrikeAI raise the
stakes further. Already linked to a campaign that breached hundreds of enterprise firewalls in 55
countries, CyberStrikeAI bundles more than 100 offensive tools into a single orchestration platform
that can automate entire attack chains with minimal human involvement. The target is always the same:
credentials, access and the systems and data behind them.

1. Gartner Inc., Emerging Tech: AI Vendor Race: Roundup For Agentic AI, Aakanksha Bansal, Danielle Casey, Alfredo Ramirez IV,
Akhil Singh, Anushree Verma, 6 October 2025 GARTNER is a trademark of Gartner, Inc. and/or its affiliates.

Why is agentic AI security different

Human users typically access a limited number of systems and perform actions one step at a time. AI agents can access multiple systems simultaneously, invoke other agents and execute complex workflows at machine speed. They often inherit privileges from user accounts or shared service accounts, giving them access that exceeds the requirements of a specific task. These characteristics require a different approach to identity security, one built around continuous visibility, risk assessment and runtime control.

How agentic AI security works

Comprehensive agentic AI security solutions discover, assess and govern AI agents across the enterprise by extending proven identity practices to agentic systems. They provide visibility into AI agents and their privileges, continuously evaluate the risks agents pose and apply adaptive policy-based controls to determine what agents can access and do. Together, these capabilities help organizations enforce least privilege and extend zero trust to AI across multicloud and hybrid environments.

Effective agentic AI security is built on three capabilities: visibility into where agents are deployed and what they can access, continuous posture assessment to identify and prioritize risk and runtime controls that enforce what agents can do and ensure credentials never reach them directly.

Visibility: Discover AI agents and their access

AI agents are proliferating rapidly. They are being deployed in end-user endpoints, SaaS solutions, cloud platforms and line-of-business workflows. Most of them operate outside established identity and access controls. They inherit privileges from users’ credentials or a shared service account.

Many enterprise security teams cannot track AI agents or their permissions. And because agents spawn dynamically, keeping inventory current requires a different approach than traditional point-in-time methods.

Agentic AI security solutions:

Discover agents across endpoints, applications and systems.
Show what systems each agent can access and what privileges it holds
Maintain a running inventory across your environment.
Identify the individual or team responsible for each agent

Posture: Assess risk and prioritize exposure

Not all AI agents present the same level of risk. Some operate within narrowly defined workflows and have access to a limited set of resources. Others can access sensitive data, interact with core enterprise applications
and control critical systems.

AI agent risks can change over time. Unlike human users or scripted processes, AI agents are  non-deterministic. Their behavior adjusts based on context, the tasks they are asked to perform, and available information.

Agentic AI security solutions:

Flag agents with privileges or access that exceed requirements.
Identify agents with inherited access to sensitive systems.
Score and prioritize exposure to focus on remediation efforts.
Update posture evaluations as agent behavior and context change.

Control: Enforce access at runtime

Understanding which AI agents exist and which agents present risk is only part of the challenge. Organizations also need a way to control agents' access to privileged resources.  

Traditional access control solutions enforce privileges when a session begins and throughout its lifecycle. Agentic AI security solutions continuously authorize AI agents at runtime, using administrator defined policies and real-time contextual information to evaluate every privileged action as it occurs.  

Agentic AI security solutions:

Enforce access decisions at the moment of each interaction.
Issue ephemeral, just-in-time credentials that expire after use.
Eliminate standing privileges that create unnecessary exposure.
Log every privileged action with a full audit trail.

Solution/concept	Function	Agentic AI security value add
Privileged Access Management (PAM)	Controls how privileged human and machine identities access sensitive systems.	Eliminates standing access privileges and provides adaptive, policy-based authorization for AI agents.
Zero Standing Privilege (ZSP)	Eliminates persistent access rights by requiring all access to be granted just-in-time and revoked immediately after use.	Enforces ZSP for AI agents at runtime, ensuring agents never hold credentials between sessions and cannot accumulate standing access over time.
Identity Governance and Administration (IGA)	Manages the identity lifecycle: provisioning, certification and policy for known identities.	Establishes ownership and accountability for AI agents, extending existing governance processes to cover agentic systems.
Non-human identity (NHI) security	Governs service accounts, API keys, secrets and machine identities.	Conceals the credentials and secrets AI agents rely on to access systems and data.
AI safety	Evaluates whether AI model outputs are accurate, fair and aligned with intended behavior.	Records and analyzes AI agent sessions to verify agents are acting within defined parameters and flag behavior that deviates from policy.

Faster to deploy: Easier to use

Delinea is consistently recognized for requiring fewer resources to manage and less time to achieve full functionality.

• 99.995% uptime SLA
• No multi-year commitment required to start

Zero standing privilege—available now

Delinea ships ephemeral access with proxy injection, JIT entitlement, and full session recording for human, machine, and AI agent identities - today.

• Native tools, broker invisible
• Time to value in weeks

Identity security built for the AI era

Delinea centralizes authorization with runtime enforcement across every AI agent in your stack.

• MCP-native connectivity
• Customers are using this in production today

Identity & access for AI agents

How do you assign and manage identities for AI agents that need to access enterprise systems?

AI agents operate independently of traditional identity and access management systems and practices. They typically inherit access privileges from user credentials or shared service accounts. Agentic AI security solutions automatically discover AI agents and dynamically control their access privileges based on administratively defined policies, giving security teams a complete, continuously updated inventory of every agent and the access it holds.

What's the difference between human privileged access and AI agent privileged access,
and why does it matter?

AI agents are autonomous and non-deterministic. Unlike humans, they operate at machine speed and can interact with many systems simultaneously. Traditional PAM solutions are designed to control human users and conventional machine identities, not systems capable of making decisions and triggering actions across dozens of resources in seconds. Controlling AI agent access requires continuous runtime enforcement, not the periodic reviews and static policies that work for human identities.

What's the difference between AI agent security and agentic AI security?

AI agent security typically refers to protecting the AI agent itself. It includes eliminating software vulnerabilities, reducing harmful outputs, preventing misuse and other agent-related security concerns. Agentic AI security, sometimes referred to as autonomous AI security, focuses on controlling what an AI agent can access and do across the enterprise. The two disciplines are complementary. Both are fundamental for ensuring AI safety, security and trust.

Control and governance

How do you enforce least-privilege principles when AI agents perform dynamic, unpredictable tasks?

Agentic AI security solutions are designed to be adaptive. They remove standing privileges and continuously authorize AI agents at runtime based on human-defined policies and real-time contextual data. They verify that the right AI agents have the right access to the right resources at the right time.

How can security teams maintain oversight and control over AI agents operating autonomously?

Agentic AI security solutions automatically discover AI agents across end user endpoints, cloud and on-premises infrastructure and SaaS platforms. They show you where those agents are deployed and what resources they can access, enabling security teams to implement adaptive, policy-based access controls with detailed logging of every privileged action.

What policies and guardrails should govern what AI agents are permitted to do within a network?

Start by eliminating standing privileges. Define which systems AI agents can access, what data they can use and what actions they can perform. Establish ownership, approval processes and monitoring requirements for higher-risk activities that deviate from standard business policies. Regularly review and update policies to accommodate new workflows, evolving business requirements and emerging threats.

Risk & threat vectors

What are the biggest security risks introduced by agentic AI systems in enterprise environments?

Most AI agents operate outside established identity and access controls, with no authorization enforcement and no audit trail. They typically inherit access rights from a user's credentials or from a shared service account with far more privileges than any single task requires. Worse still, AI agents operate at machine speed, making it difficult for security teams to identify and mitigate threats before they spread across the enterprise.

How do you detect and respond to a compromised or manipulated AI agent?

The first sign is often unexpected behavior. An AI agent may begin accessing systems it doesn't normally use, requesting unusual permissions, or performing actions outside its normal scope. Agentic AI security solutions continuously monitor agent behavior and flag deviations in real time. When an anomaly is detected, security teams can immediately restrict or revoke the agent's access, contain the session and review the full audit trail to determine the scope of the incident.

What is Claude Mythos and why does it matter for enterprise security?

Claude Mythos is Anthropic's frontier AI model, publicly disclosed in April 2026. It is highly autonomous, capable of complex multi-step reasoning and able to interact with enterprise systems at a level of sophistication that earlier models couldn't approach. Anthropic's own security research confirmed Mythos can autonomously identify zero-day vulnerabilities across major operating systems and browsers, generate working exploits and chain multiple vulnerabilities into a single exploit with minimal human involvement. For security teams, this demonstrates why runtime control over AI agents is so critical. Advanced models like Mythos operate at speeds no human can monitor, taking paths a human analyst might not anticipate.

The question is not whether AI will be used in attacks, but whether enterprise defenses are built to match the speed and autonomy of the threat. Claude Fable 5, Anthropic's release of Mythos-class capabilities, is ready for enterprise customers but is on hold pending a U.S. government order. As similar capabilities diffuse into open-source models without commercial safeguards, runtime control over AI agent access becomes a baseline requirement, not an advanced security practice.

Read the whitepaper

Secrets & credential management

How should secrets, API keys and credentials used by AI agents be stored, rotated and audited?

Secrets and other long-lived credentials should never be stored directly in agents. Agentic AI security solutions store secrets in a centralized vault, inject them at access time and rotate them automatically on a defined schedule or immediately after use. They capture every retrieval and rotation event in a unified audit trail that captures both agent and human activity. Security teams get complete visibility into which credentials agents used, when and for what purpose, without relying on agents to self-report.

Compliance & auditability

How do you create an audit trail for actions taken by AI agents to satisfy compliance requirements?

Agentic AI security solutions automatically log every action an AI agent takes, including which resources it accessed, which tools it used and which actions it performed. They provide security and compliance teams with a complete history of agent activity, including the agent's identity and owner, when the activity occurred and the outcome.

Architecture & implementation

How do you integrate agentic AI security controls into an existing privileged access management or zero trust architecture?

Agentic AI security solutions are designed to extend existing PAM solutions and zero trust architectures, not replace them. They apply the same principles already controlling human access (least privilege, just-in-time access, continuous authorization, detailed audit logging, etc.) to AI agents operating across the infrastructure. AI agents connect through existing identity providers, use the same policy engine that controls human access and feed activity into the same audit trail. There is consistent control across human and non-human identities without having to rebuild what is already in place.

What is agentic AI security?

Identity security vendors – compare the differences