Identity security is undergoing a profound evolution. While organizations historically spent years deploying human-centric identity access and control due to fragmented capabilities—ranging from authentication and authorization to privileged access and threat detection—agentic frameworks are now turning that traditional approach on its head.

Once relegated to a secondary status among cybersecurity solutions often criticized for introducing user friction, identity security is rapidly transforming into the primary control plane for autonomous agent operations. A fitting parallel was drawn during a keynote by Waymo pioneer Sebastian Thrun at SailPoint Navigate 2026, where he noted that autonomous vehicles earn rider trust not through an emergency brake, but because every decision is continuously observed, governed, and accounted for long before a vehicle ever stops.

Autonomous agents operate on the exact same principle. At major industry gatherings this year—including Cisco Live, Microsoft’s Project Perception launch, Okta Showcase and Oktane, CyberArk Impact, and SailPoint Navigate—enterprises faced a shared, pressing dilemma: how to safely hand the operational keys to AI agents. Fortunately, cybersecurity infrastructure providers are swiftly upgrading their roadmaps to manage both human and non-human identities securely.

AI Agents Raise The Stakes For Enterprise Security

The deployment of artificial intelligence agents significantly elevates enterprise risk. As businesses aggressively adopt agentic frameworks to boost productivity, they must accurately provision agents, configure API and tool access, and oversee comprehensive governance and protection. These safeguards are vital as threat surfaces expand alongside the proliferation of agents, and as malicious actors transition from AI-assisted tactics to completely autonomous attack campaigns.

The prospective weaponization of frontier AI models is another serious anxiety, highlighted earlier this year by Anthropic’s Project Glasswing and its Claude Mythos Preview, which scans massive codebases at machine speed to uncover vulnerabilities. While this serves as a defensive moat, attackers are undoubtedly deploying similar models to find soft targets and exploit zero-day flaws autonomously. Identity security vendors are answering this challenge by leaning into their core strengths and collaborating on initiatives like Project Glasswing to refine defensive capabilities.

Cisco is making a massive bet on treating the network as a security sensor and control point. During Cisco Live, CEO Chuck Robbins emphasized that the network is vastly more powerful than a standalone node. By leveraging deep network telemetry—anchored by its massive Splunk acquisition—Cisco aims to discover shadow AI, resolve connectivity bottlenecks, and enhance Duo Agentic Identity, which links every autonomous agent to an accountable human owner.

Furthermore, Cisco became Teleport’s largest strategic investor to replace shared administrator passwords with temporary access tied to enterprise networking and Duo identity controls. Building on this, Cisco introduced Privileged Infrastructure Access under a strict philosophy: if no work is occurring, no privilege should exist. While Cisco moves at machine speed, its primary challenge remains whether customers and channel partners can quickly absorb and deploy such a vast volume of offerings.

Microsoft CEO Satya Nadella recently advised companies to treat AI models as potential insider threats, keeping governance controls strictly outside the models themselves. Updates to Agent 365 and Project Perception ensure that agents receive distinct identities and per-task access rights. Should an agent exhibit behavioral drift, Microsoft’s system can restrict its capabilities, demand human approval, or execute a complete shutdown.

Okta has similarly advanced its agentic strategy, addressing core questions regarding agent location, connectivity, and capabilities. At Oktane 2026, Okta launched the Blueprint Alliance to foster an open ecosystem framework for agent security while introducing Agent Single Sign-On to substitute permanent agent passwords with temporary credentials at no additional cost. Although Okta has built a robust agentic security stack, it must continually prove its value beyond mere cloud-neutral identity orchestration.

Palo Alto Networks made a monumental identity investment by finalizing its $25 billion CyberArk acquisition, subsequently launching Idira to unify governance and access controls for people and machines. With Idira agent identity security now generally available within Prisma AIRS, customers gain critical visibility over all agents, including unmanaged ones. While entering the identity space introduces short-term friction, Palo Alto’s massive customer base provides an immediate distribution advantage.

SailPoint argues that governance remains the bedrock of agentic security. At Navigate, leadership emphasized that emergency kill switches merely treat the symptom—rogue agents—rather than the root cause: poor visibility and governance. SailPoint is expanding its Atlas intelligence layer to provide a flexible architecture supporting both human and non-human identities, deploying color-coded agents that restrict access strictly to necessary tasks.

Is Identity Security The Emerging Agentic Control Plane?

Cybersecurity infrastructure providers are aggressively positioning identity security as the definitive agentic control plane. As agentic AI graduates from pilot phases into live production, identity serves as a logical checkpoint, given that its compromise has historically driven the majority of data breaches. Organizations evaluating these tools must prioritize solutions that offer agent observability, task-based just-in-time access, continuous governance, and reliable mechanisms to neutralize rogue agents.

Ultimately, the rapid pace at which identity security vendors are innovating indicates a promising trajectory in the race to secure the future of agentic identity.