Cloud security market shifts toward runtime-first controls for AI

At a glance

What this is: The 2025 Latio Cloud Security Market Report shows cloud teams shifting from broad CNAPP consolidation toward narrower posture, exposure, and runtime controls, with AI posture management, ADR, and access management rising as priorities.

Why it matters: It matters because IAM, NHI, and autonomous-system programmes now have to govern access and runtime behaviour together, not as separate controls with separate owners.

By the numbers:

• Practitioners’ top priorities for 2026 include AI Posture Management at 65%, Application Detection & Response at 53%, Access Management at 47%, and Remediation Assistants at 35%.
• 70% of organisations grant AI systems more access than they would give a human employee performing the exact same job.
• Only 13% of organisations feel extremely prepared for the reality of agentic AI despite the majority racing toward autonomous adoption.

👉 Read Cyera's 2025 Latio Cloud Security Market Report on cloud security priorities

Context

Cloud security programmes are moving from platform consolidation toward control separation: posture tooling, exposure management, and runtime detection are being treated as distinct operating problems rather than one CNAPP buying decision. That shift matters for identity because access policy, workload identity, and data protection fail in different ways when applications are changing continuously.

For IAM and NHI teams, the report is a signal that visibility alone no longer closes the gap. The question is increasingly how quickly teams can detect misuse, constrain privilege, and respond at the pace of cloud application behaviour rather than at the cadence of periodic review.

Key questions

Q: How should security teams decide between posture, exposure, and runtime controls?

A: Use posture controls to validate baseline configuration, exposure management to identify reachable risk, and runtime controls to detect active misuse. If one control plane cannot answer all three questions, it is not sufficient on its own. Mature programmes treat them as separate layers with separate owners and response triggers.

Q: Why do AI systems need access management, not just cloud security monitoring?

A: Because AI systems can read data, invoke tools, and influence workflows, which means their access scope directly shapes business risk. Monitoring tells you what happened. Access management defines what the system was allowed to do in the first place, and that boundary must be explicit for every AI-enabled workflow.

Q: What breaks when cloud security tools only focus on scan-time posture?

A: You miss the moment when an approved configuration becomes risky during live execution. Services can drift, identities can be misused, and data paths can widen after deployment. Scan-time posture is necessary, but it does not show whether an application is behaving safely under real workload conditions.

Q: How do security teams know if runtime protection is actually working?

A: Look for evidence that suspicious behaviour is detected fast enough to contain it before the session or workload expands the blast radius. Effective runtime protection produces actionable alerts, ties them to containment steps, and shows that abnormal access can be limited during active execution, not only reviewed afterward.

Technical breakdown

Why posture-first cloud security leaves runtime gaps

Posture-first controls tell you whether a configuration is acceptable at scan time, but they do not prove the workload stays safe once it starts interacting with APIs, identities, and data paths. That is why CTEM and CADR are gaining ground: they move the control point closer to execution, where misuse, drift, and privilege abuse actually happen. In cloud environments, the security problem is often not absence of policy but the mismatch between what was approved and what runtime behaviour becomes. The same entitlement can be low-risk in one context and exposure-heavy in another, especially when data access and workload identity are coupled.

Practical implication: Treat posture findings as baseline hygiene, then add runtime telemetry where identities can reach sensitive data or privileged APIs.

Application detection and response for cloud workloads

Application detection and response focuses on identifying suspicious behaviour in application and workload flows, including unusual data access, API abuse, and escalation patterns that traditional infrastructure controls may miss. Unlike vulnerability management, which asks whether something is exploitable in theory, ADR asks whether the application is behaving in ways that indicate active risk. In cloud architectures, this is especially important where identity is ephemeral, services talk to services, and attack paths emerge through normal automation. The operational value is not just detection, but deciding what to contain before a runtime event becomes a data event.

Practical implication: Instrument applications for behaviour-based alerts and tie those alerts to response playbooks that can revoke access or isolate workloads.

AI posture management as an identity problem

AI posture management is increasingly an identity governance issue because AI systems inherit access, decision influence, and data reach from the environments they are placed in. If an AI system can read, recommend, or trigger actions across cloud and SaaS stacks, then its effective privilege must be governed like any other non-human identity. The report’s emphasis on AI posture reflects a broader reality: teams are no longer just securing compute and data planes, they are deciding what an AI system is allowed to know, touch, and act upon. That makes entitlements, usage boundaries, and oversight part of the control surface.

Practical implication: Map AI system access like NHI access, then review whether the scope matches the system's actual decision and data needs.

Breaches seen in the wild

• Azure Key Vault privilege escalation exposure — Azure Key Vault Contributor role misconfiguration enabled privilege escalation.
• Cisco DevHub NHI breach — IntelBroker exploited exposed Cisco credentials, API tokens and keys in DevHub.

Read our 52 NHI Breaches Analysis report for a comprehensive view of breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Cloud security is splitting into posture, exposure, and runtime governance because a single control plane no longer matches how modern applications fail. The report shows teams moving beyond all-in-one CNAPP thinking toward AST, CTEM, and CADR because the failure modes are different. Posture answers what is configured, CTEM answers what is exposed, and CADR answers what is happening now. Practitioners should stop assuming one platform category can govern all three.

AI posture management is becoming an identity question, not just a cloud security question. Once AI systems can access data, recommend actions, or trigger workflows, their access scope must be treated as governed entitlement rather than ambient platform capability. That aligns with OWASP-NHI and ZT-NIST-207 thinking, where access is always contextual and bounded. The implication is that teams must evaluate AI access by decision scope, not just by infrastructure location.

Runtime security is replacing static confidence as the real measure of control maturity. The market is signalling that scan-time visibility and dashboard confidence do not address workload behaviour under live conditions. What matters is whether a control can detect abnormal access, isolate misuse, and limit blast radius while the application is active. Practitioners should measure whether their current stack can respond inside the execution window, not only after the fact.

Agent-based protection is the right directional concept, but the deeper issue is identity-to-action linkage. If a workload or AI system can act on data at runtime, then the relevant governance question is not where it sits but what actions its identity can take and under what conditions. That is where NHI, access management, and response tooling converge. Security leaders should re-evaluate whether they are governing the actor or merely the platform around it.

Named concept: runtime identity drift. This report describes a market shift where privileges, exposures, and protections are no longer fixed at procurement time because cloud and AI workloads change continuously. The practical consequence is that identity boundaries must be evaluated as living states, not static assignments. Practitioners should use this lens when deciding which controls need continuous verification versus periodic review.

From our research:

• 70% of organisations grant AI systems more access than they would give a human employee performing the exact same job, according to the 2026 Infrastructure Identity Survey.
• That same survey found only 44% of organisations have implemented any policies to manage their AI agents, despite 92% agreeing that governing AI agents is critical to enterprise security.
• For lifecycle and entitlement cleanup, see NHI Lifecycle Management Guide, which helps teams tie access scope to provisioning, review, rotation, and offboarding decisions.

What this signals

Runtime security is becoming the default identity problem for cloud and AI operations. With 53% of security leaders expecting AI to run major portions of infrastructure autonomously within three years, posture-only thinking will not be enough for teams that need to govern live execution paths and access scope.

Agent-based protection is most useful when it closes the gap between entitlement and behaviour. The organisations that will feel pressure first are those whose AI and workload identities can act faster than their review cycles. That is why the governance conversation is shifting from approved access lists to observed runtime evidence.

Zero Trust only holds if identity decisions are revisited in context. When cloud and AI systems can act continuously, the programme has to distinguish between what was authorised at design time and what is safe at execution time. For a governance baseline, teams should compare these controls with NIST Cybersecurity Framework 2.0 and their own runtime response maturity.

For practitioners

• Separate posture, exposure, and runtime ownership Assign different operational owners for scan-time configuration, continuous exposure management, and live detection so each control failure has a clear response path.
• Review AI system access as non-human identity entitlement Document what data, APIs, and automation rights each AI system can touch, then compare that scope to the minimum access needed for its task.
• Add runtime telemetry to cloud response playbooks Link application and workload alerts to actions that can revoke tokens, isolate workloads, or halt risky execution before the session expands into broader impact.
• Use CTEM to validate actual exposure paths Prioritise the assets and identities that can be reached from current cloud and application paths, rather than relying only on planned segmentation or approved architecture.
• Align access reviews to behaviour, not only entitlements Combine periodic entitlement review with evidence of real usage so stale permissions and over-broad runtime access do not look safe just because they were approved.

Key takeaways

• The report shows cloud security moving away from single-platform consolidation toward distinct posture, exposure, and runtime controls.
• AI posture management is increasingly an identity governance issue because AI systems can inherit access that outpaces human-equivalent privilege decisions.
• Practitioners should judge control maturity by runtime containment and behavioural visibility, not by scan-time confidence alone.

Key terms

• Cloud Application Detection And Response: Cloud Application Detection and Response is the practice of identifying suspicious behaviour inside live cloud applications and workloads, then containing that behaviour before it spreads. It focuses on runtime signals such as unusual access, API misuse, and escalation paths that scan-time posture tools cannot see.
• AI Posture Management: AI Posture Management is the control discipline that evaluates how AI systems are configured, what they can access, and whether those permissions match the task they are meant to perform. In practice, it treats AI systems as governed identities with measurable entitlement boundaries.
• Continuous Threat Exposure Management: Continuous Threat Exposure Management is the ongoing process of finding which assets, identities, and paths are actually reachable from the current environment. It moves risk assessment away from static inventories and toward live exposure, so security teams can prioritise what an attacker or misuse path can reach now.
• Runtime Security: Runtime security is protection that operates while applications or workloads are actively executing, rather than only before deployment. It matters because cloud and AI systems can drift, invoke tools, or expose data in ways that approved configurations do not predict.

Deepen your knowledge

Cloud posture, exposure management, and runtime detection are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are aligning identity governance to cloud and AI workloads, it is worth exploring.

This post draws on content published by Cyera: 2025 Latio Cloud Security Market Report. Read the original.