Multicloud security in 2026: Securing the rise of autonomous identities

Multicloud is now the enterprise reality. 88% of organizations operate in hybrid or multicloud environments, while 81% use two or more cloud providers for critical workloads (Source: 2026 State of Cloud Security Report) . The security challenge has therefore shifted from protecting individual cloud environments to understanding how identities, workloads, data and permissions interact across them.

A bigger risk lies in what connects these environments and what moves between them.

From security tools to a connected security platform

Cloud security did not start with a single platform. It evolved through a collection of specialized capabilities. Organizations deployed CSPM to assess cloud posture, CWPP to protect workloads, CIEM to manage permissions, KSPM to secure Kubernetes, DSPM to protect data and, more recently, AI-SPM to address the security risks associated with AI workloads and applications. Each tool addressed a genuine need. The problem emerged when those tools operated independently. An attacker does not move through an enterprise by product category. A single attack path can begin with a cloud misconfiguration, move through an over-privileged identity, reach a workload and ultimately expose sensitive data. The individual controls may all be working as designed, yet security teams can still miss the relationship between them. That is the gap the industry is now trying to close.

The answer has been cloud native Application Protection Platform (CNAPP). Rather than treating posture, workload, identity, data and application security as separate disciplines, CNAPP brings them together to create a more connected view of cloud risk. The industry's direction is increasingly clear. Google's approximately $32 billion acquisition of Wiz underscored the strategic importance of cloud native security consolidation. At the same time, Microsoft continues to position Defender for Cloud as a unified security platform across Azure, AWS and Google Cloud. Palo Alto Networks has similarly evolved Prisma Cloud into Cortex Cloud, bringing cloud security together with cloud detection and response. The details differ by vendor, but the strategic shift is the same:

Cloud security is moving from a collection of point products toward connected platforms that can understand risks across the cloud lifecycle.

CNAPP is becoming the foundation. But it is not the end of the story.

The cloud now has a new kind of user

Just as cloud security platforms are consolidating, the nature of activity inside the cloud is changing.

AI agents are becoming active participants in enterprise environments.

Unlike traditional applications or human users, agents can interpret information, make decisions and take actions. They can read emails, query databases, invoke APIs, create resources and interact with applications, often with limited human intervention. This introduces a fundamental identity challenge. Enterprise identity systems were largely designed around people and relatively predictable machine accounts. Agentic systems introduce identities that can be created quickly, operate autonomously, access multiple services and potentially disappear once their task is complete. The scale is also different. Non-human identities can now outnumber human identities by roughly 50:1 in enterprise environments, though the ratio varies considerably across organizations and architectures. This means organizations are no longer securing only employees, applications and workloads, but they are securing autonomous actors. And those actors can cross cloud boundaries.

The warning signs are already visible. SailPoint research found that 80% of organizations using AI agents reported that those agents had taken unintended actions, including accessing systems they were not supposed to access or sharing inappropriate data. Meanwhile, a 2026 Cloud Security Alliance study found that 68% of organizations cannot clearly distinguish AI-agent activity from human activity. The same study found that 74% said agents often receive more access than necessary, while 79% believe agents create new access pathways that are difficult to monitor. These are not simply AI governance issues.

They are identity security issues.

Identity is becoming the control plane

This is why identity is emerging as the new security perimeter. Cloud environments are increasingly accessed through identities rather than traditional network boundaries. Permissions determine which workloads can communicate, which applications can access data, which APIs can be invoked and which actions an agent can take. When those identities span multiple clouds, a compromised credential or excessive permission can provide a pathway across the enterprise. The challenge becomes even greater when the identity is non-human. An employee typically has an established role, manager, business function and lifecycle. An AI agent may be created for a specific task, given access to several systems and expected to act autonomously. Its permissions, therefore, need to be evaluated in context and not simply granted once and forgotten. This requires organizations to rethink identity governance. Every agent should have a clear owner, defined purpose, appropriate permissions and an auditable lifecycle. Access should be limited to what is necessary, granted for as long as necessary and continuously evaluated as circumstances change. In other words:

AI agents need identity governance just as seriously as human users do, but the controls must be designed for machine-speed activity.

Microsoft's investments in agent management, identity and security reflect this broader shift toward securing AI agents as they become integrated in enterprise workflows.

One compromised identity can still change the story

The importance of identity is not theoretical. In July 2026, Bank of Baroda disclosed that a data leak involved the compromise of a single employee's email account, stating that its core banking systems were not accessed and remained secure. Reports about the incident alleged that up to 1TB of data, including customer information, had been exposed; the bank said it had initiated a forensic investigation and notified relevant authorities. The significance of cloud security is not the size of the reported dataset alone. It is the entry point. The incident is a reminder that an attacker does not necessarily need to compromise the most sophisticated part of an organization's infrastructure. One identity, with the wrong access at the wrong time, can become the starting point for a much larger security problem.

The security gap is now widening

Apart from identity, security teams are also dealing with three converging challenges:

• Tool sprawl: The 2026 Cloud Security Report found that 69% of organizations identify tool sprawl and visibility gaps as a top barrier to effective cloud security. More technology does not automatically create more visibility. Without correlation, it can create more noise.
• Skills gap: The same report found that 74% of organizations report a shortage of qualified cybersecurity professionals, while 59% remain in the early stages of cloud security maturity.
• Machine-speed attacks: Attackers are using automation and AI to discover vulnerabilities, exposed services, credentials and misconfigurations faster than traditional security operations can respond. 66% of organizations lack strong confidence in their ability to detect and respond to cloud threats in real time.

That creates a fundamental imbalance; the environment operates at machine speed. The attacker operates at machine speed. The defender cannot remain dependent on manual processes.

From visibility to autonomous security

The answer is not another dashboard. For years, security teams have focused on collecting more telemetry and generating more alerts. The next stage is about turning that information into context, prioritization and action. This is where the evolution of CNAPP becomes particularly important. The modern cloud security platform needs to connect signals across code, cloud configuration, identity, workloads, data and runtime activity. More importantly, it needs to understand how those signals relate to one another. A misconfiguration by itself may be low risk, an over-privileged identity by itself may also appear manageable and an exposed database may generate another alert. But combine those three conditions and the risk can change dramatically. The security platform, therefore, needs to understand attack paths, identity context and business impact and not simply individual alerts. And increasingly, it needs to respond automatically. The goal is not to see more.

The goal is to understand more, prioritize faster and act before the attack path becomes an incident.

The question CISOs should be asking in 2026

This changes the conversation around multicloud security. Instead of asking:

“How do I secure my clouds?”

The more important question is:

"Who and what is moving across my cloud environments, what are they authorized to do and would I know if one of them suddenly behaved differently?"

Multicloud security is no longer simply about putting stronger controls around AWS, Azure or Google Cloud individually. Those environments are increasingly connected and security platforms are increasingly bringing their controls together. The harder challenge is understanding the identity and activity layer connecting them. It includes the employee whose credentials span multiple platforms. The service account with permissions nobody has reviewed recently. The API key has quietly become a permanent pathway into an application. The workload is moving between environments. And increasingly, the AI agents can make decisions and take actions without waiting for a human. The attack surface has therefore become more dynamic and much harder to define.

Four priorities for enterprises in 2026

• Treat CNAPP as the foundation, not the destination: Consolidation creates the visibility and context required to understand risk across cloud environments. But organizations should use that foundation to move toward continuous detection, prioritization and response.
• Build an inventory of non-human identities: Organizations cannot govern what they cannot see. Service accounts, workload identities, API keys and AI agents need clear ownership, purpose and lifecycle controls.
• Make identity a primary security control point: Least privilege, just-in-time access, contextual authorization and continuous verification should apply to humans, workloads and agents alike.
• Match the attacker's speed: Automated threats require automated detection and response. Security teams should increasingly allow platforms to contain, revoke, isolate or remediate clearly defined risks without waiting for manual intervention.

The future of multicloud security will therefore be less about building higher walls around individual clouds and more about establishing continuous trust across everything that connects and moves between them. Multicloud security is no longer about securing each cloud in isolation, but it is about securing the identities, workloads, data and AI agents that move between them.