Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By Justin Dolly, Chief Customer and Security Officer, Ory Corp.
As we look toward 2026, one prediction stands out from all the others: autonomous AI agents will fundamentally break traditional identity and access management systems, forcing enterprises to rethink their security infrastructure.
This isn’t speculation. The transformation is already underway. In recent research by Ory and Enterprise Management Associates (EMA), “Agentic AI Identities – Is Your Organization Prepared?“, large enterprises have deployed AI agents at scale, with 83% of organizations with more than 1,000 employees running agents in production environments today. These aren’t simple chatbots or assistive tools; rather, they are autonomous systems that write and deploy code, manage cloud infrastructure, execute financial trades, and access both internal company data and sensitive customer information.
We’re staring down a big problem. The identity infrastructure supporting these deployments was never designed for this agentic reality.
The Identity Crisis Nobody Saw Coming
Traditional IAM systems were built around a fundamental assumption: identities represent humans. Every authentication flow, every authorization policy, every security control was designed with human behavior patterns in mind. For example:
- Rate limiting? Based on human typing speeds.
- Access approvals? Managing human “in-the-loop”.
- Risk scoring? Calibrated against human attack patterns.
The new reality is AI agents obliterate these assumptions.
When an agent can execute 10,000 API calls per second, authenticate across hundreds of services simultaneously, and operate 24/7 without rest, traditional IAM systems don’t just slow down and become sluggish, they risk entire collapse.
We see this firsthand as organizations struggle to distinguish between legitimate agent activity and sophisticated attacks. Old rules and solutions simply don’t apply.
2026 will be an inflection point with agents outnumbering human users by a factor of three or more in most enterprise environments. This isn’t a gradual shift that gives IT teams time to adapt. It’s a tsunami that requires immediate architectural rethinking.
Three Predictions That Will Define 2026
First, we’ll see the emergence of agent-native IAM as a distinct category. Any modular strategy remains valid but forcing agent identities into human-focused IAM systems will fail spectacularly. By mid-2026, forward thinking enterprises will separate agent identity management into dedicated systems designed specifically for machine-to-machine authentication at massive scale. These systems will use fundamentally different approaches to session management, token lifecycle, and access control that reflect how agents actually operate rather than forcing them into human-shaped boxes.
Second, the “deployment gap” will widen before it narrows. That same “preparedness” study shows that only 55% of organizations with deployed agents have documented policies governing agent behavior. Even more concerning, the majority of enterprises admit their current IAM infrastructure cannot handle the authentication and authorization load that agents generate, yet they’ve deployed them anyway. We’ve already seen the consequences. In summer 2025, an AI coding agent deleted a company’s entire production database, then attempted to cover it up. Just recently, a security researcher discovered that a billion-dollar legal AI tool exposed over 100,000 confidential files due to inadequate access controls.
Throughout 2026, these incidents will escalate in frequency and severity as more organizations deploy agents without proper governance. These expensive wake-up calls will ultimately accelerate the shift toward proper governance of agent IAM by year’s end.
Third, regulatory frameworks will begin treating agent identities as a distinct compliance concern. When agents have direct access to customer data, financial systems, and intellectual property, the compliance implications become impossible to ignore. Expect new guidance from bodies like NIST and industry-specific regulators requiring organizations to demonstrate that they can authenticate agent identities, authorize agent actions, and maintain detailed audit trails of agent behavior.
Organizations that haven’t invested in proper agentic IAM will find themselves scrambling to meet these emerging requirements.
What This Means for the Industry
The shift to agentic IAM will reshape vendor landscapes throughout 2026. Legacy IAM providers built on monolithic SaaS architectures will struggle to adapt. Their licensing models, designed around per-user pricing, break down when agent populations explode. Their deployment models, centered on cloud-only offerings, clash with enterprise requirements (and trending preferences) for on-premises or private cloud deployments, particularly for highly regulated industries.
The winners in this space will share several characteristics. They’ll offer genuine deployment flexibility, letting customers choose between self-hosted and SaaS options rather than forcing a single model. They’ll build on standards-based approaches because custom integration for every agent type quickly becomes untenable at scale. They’ll design API-first architectures from the ground up since agents interact with systems fundamentally differently than humans clicking through web interfaces. At Ory, we built our solutions with these principles as native, foundational, not as afterthoughts. Organizations that deployed agents with minimal governance six months ago are now desperately trying to retrofit proper identity controls. The ones getting it right are treating agent IAM as a greenfield problem rather than an incremental upgrade to existing systems.
The Path Forward
There is a path forward. 2026 will also be the year we develop practical patterns for managing agent identities at scale. Organizations will learn to implement fine-grained authorization that goes beyond simple role-based access control to consider context, risk levels, and agent behavior patterns. They’ll build observability systems that can actually make sense of agent activity logs when a single agent might generate millions of events per day. They’ll develop governance frameworks that define clear boundaries for what agents can and cannot do, with technical enforcement rather than policy documents that nobody reads.
The challenge for technology leaders is recognizing that this transformation isn’t optional. You can’t wait for perfect solutions to emerge. The agents are already deployed, already accessing your systems, and already creating risk. The organizations that thrive in 2026 will be those that acknowledge the identity crisis that agents have created and take concrete steps to address it now, even if those solutions feel imperfect.
This is the year agent identity moves from an emerging concern to a board-level priority. The question isn’t whether your organization will face this reckoning, but whether you’ll be prepared when it arrives.
A Point of Optimism
Here’s the reality that gives me optimism as we head into 2026: while current IAM systems may struggle to handle the combined load of customers, partners, employees, and now agents, the technology to solve this problem already exists. We know how to authenticate agent identities. We know how to authorize and control their actions. The architectural patterns are proven, and the standards are established.
The shift is achievable. Organizations that act decisively in 2026 and invest in modern, flexible IAM infrastructure built for machine-scale operations that implement proper governance before the next headline incident, these are the ones that will harness AI agents as a competitive advantage rather than a liability.
If we approach this transformation with the seriousness it deserves and build the right foundations now, then AI and agents can change how we work for the better rather than creating the security nightmare everyone fears.
The tools are ready. 2026 is the year we all decide which path we’re on.
##
ABOUT THE AUTHOR
Justin Dolly is Chief Customer and Security Officer at Ory Corp, where he brings decades of global security leadership to help enterprises transition from legacy and homegrown IAM and CIAM systems to modern, trusted identity architectures. A seasoned CISO, Dolly has held senior security leadership roles at organizations including CBS Interactive/CNET Networks, ServiceNow, VMware, Malwarebytes, SecureAuth, and Sauce Labs. He is also a current or former board member and advisor to several technology companies, including CrowdStrike, SecureAuth, Smarsh and Cloudbolt. Known for building high-performing global teams and embedding security-first cultures, Dolly is a Certified Chief Information Security Officer (CCISO) with deep expertise in enterprise security, privacy, and large-scale, compliance-driven environments.
