Executive Overview
The modern enterprise is undergoing a silent, high-speed structural revolution. Artificial intelligence has decisively broken out of the sandbox. No longer confined to isolated testing environments or experimental chatbots answering basic queries, AI agents are rapidly becoming the primary execution layer of contemporary business infrastructure. These systems write code, process financial transactions, interact directly with customer databases, and execute multi-step workflows with minimal human intervention.
Yet, this operational leap has opened a massive, potentially catastrophic vulnerability gap. While corporations race to deploy autonomous agents to drive efficiency, the security controls required to govern them have languished in the analog era.
According to the latest Horizons of Identity Security report published by enterprise identity governance leader SailPoint, 79% of global enterprises already have AI agents running live in production environments. Simultaneously, a staggering only 2% have deployed dedicated identity security tools specifically designed to govern, monitor, and control these non-human actors.
This widening chasm between operational deployment and security governance represents one of the most perilous attack surfaces in the history of enterprise computing. Autonomous agents are routinely granted access to critical applications, sensitive intellectual property, high-privilege credentials, and core infrastructure. In doing so, they are effectively establishing an entirely new, highly dynamic class of digital identity that traditional IT and security architectures are entirely unequipped to manage.
In response to this existential threat, SailPoint used its flagship Navigate conference to unveil a comprehensive suite of identity security capabilities designed explicitly for the agentic era. By expanding its Agentic Fabric, introducing specialized autonomous security agents, and rolling out advanced visibility and posture management tools, SailPoint aims to transition enterprises from security chaos to rigorous operational control.
This report examines the dimensions of the agentic identity crisis, dissects SailPoint’s technological counter-offensive, analyzes executive perspectives, and explores the long-term implications of securing autonomous systems.
Detailed Chronology: The Emergence of the Agentic Security Gap
To understand the urgency behind SailPoint’s recent announcements, one must examine the rapid evolution of enterprise artificial intelligence over the past three years.
Phase 1: The Copilot Era and Static Credentials (2023–2024)
When large language models (LLMs) first entered the enterprise mainstream, they primarily functioned as static assistants or "copilots." These tools were fundamentally reactive: a human user prompted the model, and the model generated text, code suggestions, or summaries. From an identity and access management (IAM) perspective, these early models inherited the permissions of the human operator sitting at the keyboard. Security teams treated them as extensions of the user, relying on legacy role-based access control (RBAC) models and standard OAuth tokens.
Phase 2: The Proliferation of Shadow AI and Tool-Calling Agents (2024–2025)
As artificial intelligence architectures matured, developers shifted from building conversational models to deploying agentic workflows. These systems were endowed with agency—the ability to plan, execute, and evaluate tasks autonomously. Equipped with "tool-calling" capabilities, modern AI agents can execute APIs, query databases, spin up cloud instances, and dispatch emails on behalf of the organization.
However, because these agents were often spun up rapidly by line-of-business units eager for productivity gains, they bypassed standard IT procurement and security vetting processes. This led to an explosion of "shadow AI"—unsanctioned agents, Model Context Protocol (MCP) servers, hardcoded API keys, and orphaned data stores operating entirely outside the purview of Chief Information Security Officers (CISOs).
Phase 3: The Breaking Point and SailPoint’s Navigate 2026 Intervention (2026)
By early 2026, the velocity of agentic operations overwhelmed traditional human-centric security frameworks. When a single autonomous agent can initiate thousands of programmatic tool calls per second, human-in-the-loop approval workflows become functionally obsolete.
Recognizing that manual identity administration was buckling under the weight of autonomous automation, SailPoint utilized its Navigate 2026 conference to fundamentally redefine enterprise identity security. The company unveiled a multi-pronged architectural defense centered on three core pillars:
- The Expanded Agentic Fabric: Designed to discover shadow AI agents, establish runtime authorization, and provide an immediate "kill switch" to disable compromised or rogue agents.
- SailPoint Autonomous Agents: AI systems specifically engineered to govern other AI systems, operating at machine speed to intercept threats and adjust permissions dynamically.
- Autonomous Identity Security Posture Management (A-ISPM): A platform-driven capability leveraging SailPoint Atlas telemetry to uncover dormant accounts, shadow administrators, and risky non-human access paths.
Supporting Context & Metrics: The Scale of the Challenge
The statistical realities highlighted in SailPoint’s Horizons of Identity Security report paint a stark picture of modern enterprise vulnerability. The mismatch between AI adoption velocity and identity governance maturity is not merely an administrative oversight; it is an open invitation for sophisticated cyber exploitation.
Key Metrics from the Research
- 79%: The percentage of enterprise organizations currently running AI agents in live production environments. These agents are actively interacting with business-critical systems, customer records, and financial databases.
- 2%: The percentage of those same enterprises that have deployed purpose-built identity security tools capable of governing non-human AI identities. This 77-percentage-point deficit represents one of the widest security gaps in modern corporate history.
- 10,000+: The estimated volume of programmatic tool calls executed by advanced enterprise AI agents per second during peak operational workflows—a velocity utterly unmanageable by human review boards.
The Anatomy of Non-Human Risk
Unlike traditional human users—who log in during business hours, exhibit behavioral baselines, and possess verifiable human attributes—AI agents operate continuously, scale elastically, and make decisions based on probabilistic algorithms.
When an agent is granted access to credentials, databases, and third-party APIs, it inherits a vast attack surface. If an adversary compromises an agent’s underlying prompt context, injects malicious instructions (indirect prompt injection), or steals its API tokens, the attacker gains high-speed programmatic access to every system the agent touches. Without granular runtime authorization, identity governance, and immediate revocation mechanisms, organizations are blind to lateral movement executed at machine speed.
Official Statements and Industry Insights
The paradigm shift required to secure the agentic era was a central theme among SailPoint’s executive leadership at Navigate 2026. Industry veterans emphasized that traditional security dogmas must be discarded in favor of automated, real-time governance.
Redefining Administrative Reality
Addressing the sheer impossibility of manual oversight in an automated world, SailPoint CTO Chandra Gnanasambandam didn’t mince words during his keynote address:
"The idea that a security admin is going to review and approve access for a group of autonomous agents making 10,000 tool calls a second isn’t just outdated — it’s a fantasy."
Gnanasambandam’s assessment underscores the fundamental thesis of SailPoint’s new product architecture: human administrators can no longer serve as the gatekeepers of micro-transactions executed by software agents. Security governance must be automated, continuous, and intelligent enough to operate at the speed of the machine.
The New Execution Layer
Echoing this sentiment, SailPoint President Matt Mills emphasized the structural transformation taking place across global corporations:
"Autonomous agents are no longer just experimental copilots; they are becoming the primary execution layer of modern enterprises."
Mills highlighted that as enterprises transition from human-driven applications to agent-driven execution, identity has become the ultimate perimeter. Protecting that perimeter requires moving beyond static role assignments toward dynamic, context-aware governance frameworks.
Technical Deep-Dive: Inside SailPoint’s Agentic Defense Architecture
To address the vulnerabilities outlined in its research, SailPoint’s engineering teams constructed a comprehensive technological ecosystem designed to discover, monitor, and control autonomous entities across the enterprise.
1. Advanced Discovery and Visibility via the Agentic Fabric
You cannot secure what you cannot see. The first line of defense in SailPoint’s updated architecture is comprehensive asset discovery.
Using specialized endpoint and browser sensors, integration with Security Information and Event Management (SIEM) and Extended Detection and Response (XDR) telemetry, and advanced non-human identity discovery algorithms, the platform scans enterprise environments for unmanaged elements. This includes:
- Shadow AI Agents: Unauthorized scripts and autonomous models deployed by business units.
- Model Context Protocol (MCP) Servers: Infrastructure components that bridge AI models with enterprise data sources.
- Orphaned Credentials and API Keys: Hardcoded secrets left in code repositories or cloud configurations.
- Hidden Data Stores: Ungoverned databases and storage buckets accessed by autonomous tools.
2. Runtime Authorization and the Emergency Kill Switch
Discovery is only half the battle; organizations must also enforce real-time control. SailPoint’s expanded Agentic Fabric introduces runtime authorization capabilities. Instead of relying on static permissions granted at deployment, the fabric intercepts tool calls and data requests in real-time.
Crucially, if an agent exhibits anomalous behavior, displays signs of compromise, or attempts unauthorized data exfiltration, security teams—or automated systems—can trigger an immediate revocation and kill switch, neutralizing the rogue agent within milliseconds without disrupting stable enterprise operations.
3. AI Governing AI: SailPoint Autonomous Agents
In a brilliant application of recursive technology, SailPoint introduced SailPoint Autonomous Agents—specialized AI constructs whose sole purpose is to govern other AI systems.
These security agents continuously monitor runtime activity across the enterprise, intercept emerging threats, and dynamically adjust permissions. If a legitimate business agent requires temporary access escalation to complete a complex workflow, the security agent can evaluate the context, apply policy constraints, and grant Just-in-Time access, all without requiring human intervention.
4. Autonomous Identity Security Posture Management (A-ISPM)
Building upon SailPoint’s centralized Atlas platform, the newly introduced A-ISPM capability brings order to structural chaos. A-ISPM continuously crawls enterprise infrastructure to identify:
- Dormant Accounts and Inactive Agents: Digital identities that retain high privileges long after their business utility has expired.
- Orphaned Access Paths: Unmonitored permissions linking external tools to core databases.
- Shadow Administrators: Non-human or human entities wielding unchecked elevated privileges.
Once identified, organizations can configure automated workflows to revoke excessive access, disable compromised agents, assign formal ownership, or enforce compliance policies automatically.
5. Just-in-Time (JIT) Provisioning
Moving away from the dangerous precedent of permanent, standing privileges, SailPoint’s strategy heavily emphasizes Just-in-Time (JIT) provisioning. Humans, service accounts, and AI agents are granted access exclusively when a verified operational need arises. These permissions are bound to strict conditional parameters, such as mandatory reauthentication steps, time-bound expiration windows, and explicit business justifications.
Future Outlook: The Road Ahead for Enterprise Identity Security
The rollout of SailPoint’s agentic security capabilities—with core features beginning global deployment immediately and A-ISPM slated for release in the fourth quarter of fiscal 2027—marks a critical inflection point for the cybersecurity industry. However, the war against automated risk is only just beginning.
The Proliferation of Multi-Agent Ecosystems
As enterprises move beyond single isolated agents toward interconnected, multi-agent systems—where AI agents converse, negotiate, and delegate tasks to other AI agents independently—identity governance will face unprecedented complexity. Tracing accountability across a chain of five autonomous agents collaborating on a cross-border financial transaction will require advanced cryptographic lineage tracking and real-time behavioral attribution.
The Regulatory Imperative
Governments and regulatory bodies worldwide are rapidly waking up to the risks of autonomous systems. With frameworks like the European Union Artificial Intelligence Act taking effect and tightening data privacy mandates globally, enterprises will soon be legally required to demonstrate verifiable governance over automated decision-making systems. Failure to implement robust non-human identity controls will no longer result merely in security breaches; it will trigger severe regulatory penalties, compliance failures, and catastrophic liability.
Conclusion: Embracing the Autonomous Era Safely
The rise of agentic AI is irreversible. The productivity gains, operational velocity, and innovation unlocked by autonomous systems are too profound for enterprises to ignore. Yet, as SailPoint’s data demonstrates, operating in the agentic era without robust identity governance is akin to flying a supersonic jet without instrumentation.
By shifting from manual, human-centric oversight to machine-speed, AI-driven governance frameworks like those introduced at Navigate 2026, organizations can successfully bridge the identity gap. The future belongs not to those who lock down and reject artificial intelligence, but to those who master the art of governing autonomy—transforming potential chaos into secure, scalable enterprise control.
