Identity, trust and governance are the second floor of security

At a glance

What this is: This article argues that the CIA triad remains foundational for data protection, but modern security decisions now sit above it in identity, trust, and governance.

Why it matters: IAM, PAM, NHI, and agentic AI programmes need to govern actors and delegation chains, not just data states, because most breach paths now exploit identity and trust.

By the numbers:

• 74% of attacks start with compromised identities, according to C1.ai.
• 79% of cyberattack detections are malware-free, according to C1.ai.
• 88% of web application breaches involve stolen credentials, according to C1.ai.

👉 Read C1.ai's analysis of identity, trust and governance as security's second floor

Context

The core governance gap is that the CIA triad describes properties of data, while modern security failures increasingly arise from the actors touching that data. As infrastructure, workloads, and now AI agents become more abstracted, the decisive control point shifts from protecting information alone to governing identity, trust, and delegated authority.

C1.ai frames this as a second-floor problem rather than a fourth pillar problem. That distinction matters because it changes the programme design question from how to harden data paths to how to control who or what is acting, what it is trusted to do, and under what rules that trust is revoked or limited.

For identity teams, the article connects human IAM, non-human identities, and agentic systems into one control-plane model. The practical implication is that identity governance now has to scale across humans, workloads, service identities, and autonomous actors without assuming the same review cadence or trust model fits all three.

Key questions

Q: How should security teams govern identity as a control plane?

A: Security teams should treat identity as the layer that decides who can act, how far authority travels, and what context makes an action legitimate. That means combining IAM, PAM, and NHI lifecycle controls with continuous evaluation, machine-readable policy, and delegation tracking. If identity is only provisioned and not governed, the rest of the stack is protecting decisions that were never validated.

Q: Why do stolen credentials and compromised identities create more risk than classic perimeter failures?

A: Because most modern attacks exploit the actor, not the data store. Once an attacker has a trusted identity, they can request legitimate access, move through allowed paths, and avoid many controls that only look for malware or network intrusion.

Q: What breaks when access review does not cover non-human identities used by AI agents?

A: When access review ignores the NHIs behind AI agents, organisations lose visibility into stale privileges, inherited rights, and abandoned credentials that still allow action. That creates an audit gap and a control gap at the same time. The access path may still work even when no one can explain why it should.

Q: How do IAM, PAM, and NHI governance differ from data security controls?

A: IAM, PAM, and NHI governance answer who or what is allowed to act, under what trust conditions, and with what scope. Data security controls answer how the information is protected once that decision has already been made, so the two layers serve different purposes and must be managed separately.

Technical breakdown

Why the CIA triad stops at the data plane

Confidentiality, integrity, and availability are data-plane properties. They describe what must be true about information after access is already decided, but they do not answer who is requesting access, whether that requester is trusted, or which policy should govern the request. As environments moved from physical servers to virtual machines, containers, serverless functions, and AI agents, the security problem shifted upward. The old model still matters, but it no longer covers the full decision surface.

Practical implication: treat CIA as the protection layer and pair it with explicit identity governance for access decisions.

Identity and trust as the security control plane

A control plane decides, authorises, and coordinates. In modern security architecture, identity determines who or what is acting, trust determines how much confidence the system places in that actor right now, and governance determines the rules and accountability behind the decision. That is why identity now behaves like the control plane above CIA: it does not replace data protection, it directs it. Zero Trust models, continuous access evaluation, and machine-readable policy all exist because static perimeter logic no longer fits distributed systems.

Practical implication: design access policy, trust evaluation, and revocation as continuous controls, not one-time gate checks.

Why delegated authority changes with agentic AI

Agentic systems add a new layer of complexity because the human sponsor is no longer the only decision-maker in the chain. The article describes agents that call tools, create sub-agents, and operate at machine speed, which means delegation can expand faster than human oversight can review it. The governance problem is no longer only whether a credential exists, but whether the actor can change its own execution path while carrying inherited authority. That breaks the assumption that access can be reviewed after the fact.

Practical implication: govern agent permissions at issuance time and trace delegation chains end to end.

NHI Mgmt Group analysis

Identity, trust, and governance are not additive controls to CIA, they are the layer that decides whether CIA can still be meaningful in modern environments. The article is right to reject the idea that data properties alone can govern actor behaviour. Once the environment includes workloads, service identities, and autonomous systems, the decisive risk is no longer just data exposure but uncontrolled authority. Practitioners should treat the control plane as the primary governance problem, not a sidecar to protection.

The second-floor model is the right abstraction for non-human identity governance. Service accounts, tokens, certificates, and AI agents do not fail the same way as data, so they cannot be managed as if the same security model applies. The article’s strongest insight is that identity is a property of actors, while CIA is a property of data. Practitioners should organise NHI and workload governance around actor lifecycle and delegation, not around storage-centric controls.

Automated and autonomous systems collapse the assumption that access review is enough. Access review processes were designed for access that persists long enough to be observed, certified, and remediated. That assumption fails when an AI agent can obtain, use, and release authority inside one execution window or spawn downstream actors before a reviewer ever sees the state. The implication is not just tighter review, but a different governance model for time-bound decision authority.

Identity blast radius: The article surfaces a useful concept for modern programmes: the blast radius of a breach is increasingly determined by who or what is trusted, not by where data is stored. When delegated authority is broad, compromised credentials or over-trusted agents can move laterally across systems faster than traditional data controls can react. Practitioners should measure security exposure in terms of identity reach and delegation depth.

Governance becomes the differentiator once identity, trust, and policy execute at machine speed. The article correctly places accountability at the centre of the model because policy without traceable ownership is only automation. For IAM, PAM, and NHI programmes, that means governance must be operational, machine-readable, and auditable across every actor class. The programme question is no longer whether identity is involved, but whether the organisation can govern it continuously.

What this signals

Identity, trust, and governance should now be treated as programme-level architecture, not as supporting controls. If your security roadmap still centres on data protection alone, you are optimising the ground floor while breaches increasingly occur upstairs. The practical shift is to budget, staff, and measure the control plane with the same seriousness as the data plane.

Second-floor governance requires different operating metrics. Review cadence, policy execution speed, delegation depth, and revocation latency matter more than broad compliance checklists when actors can move at machine speed. For identity leaders, the real test is whether the organisation can govern the fastest entity in the environment without relying on human-paced oversight.

For practitioners

• Map your control plane Inventory which actor types you govern today, including humans, service identities, workloads, and AI agents, then identify where access decisions are still being made with data-centric controls only.
• Trace delegated authority end to end Document how authority moves from a human sponsor to workloads, APIs, and any agentic sub-processes so you can see where inherited trust expands beyond intended scope.
• Move access decisions to issuance time Use machine-readable policy and continuous authorisation where actors can act, delegate, or expire faster than quarterly review cycles can capture.
• Rebuild review cadences for fast actors Separate human recertification cycles from NHI and agent lifecycles so review timing matches the way each actor actually acquires and sheds authority.
• Measure identity reach, not just data exposure Track how far one compromised identity can move across systems, because delegation depth and trust propagation now shape blast radius more than perimeter placement.

Key takeaways

• The article reframes modern security as an actor-governance problem, not only a data-protection problem.
• Its main claim is that identity, trust, and governance now carry the decisions that CIA cannot make on its own.
• For practitioners, the implication is to govern humans, workloads, and AI agents as separate actor classes with different review and trust models.

Key terms

• Control Plane: The control plane is the set of actions that create, configure, or manage a service. For AI workloads, it covers deployment and administration of the model platform, while data-plane permissions govern what the service and its identities can read or process.
• Delegation Chain: A delegation chain is the sequence of identities, credentials, and tool calls an agent uses to complete a task across systems. It matters because each step may appear acceptable on its own while the combined path produces an outcome no reviewer would have approved directly.
• Actor Identity: Actor identity describes the identity of the entity making the request, not the information being protected. In modern environments that actor may be a person, workload, service account, or AI agent, each requiring a different governance model.
• Continuous authorization: Continuous authorization is the practice of rechecking access as a session unfolds instead of trusting a single login decision. It matters for AI workflows because the request, context, retrieved data, and downstream action can all change between prompt and execution, making static approval too blunt.

What's in the full article

C1.ai's full blog covers the argument and evidence this post intentionally leaves in summary form:

• The full control-plane framing that links identity, trust, and governance to Zero Trust and policy-as-code
• The infrastructure-era comparison table showing how security responsibility moved from physical servers to agentic systems
• The article's discussion of delegation chains, sub-agents, and machine-speed authority in AI environments
• The market and standards references that support the case for a second-floor security model

👉 C1.ai's full blog expands the control-plane model, delegation chain implications, and the shift from CIA-only thinking.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity security are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.