NHI & Agentic AI Identity Security FAQ

Question 1

Why do consumer accounts need different IAM controls from workforce identities?

Accepted Answer

Consumer identities operate at much larger scale, with higher volatility, more self-service, and far less direct administrator oversight. That changes the control model. B2C CIAM has to handle registration, consent, social login, and account recovery in ways that support customer journeys while still resisting fraud and takeover attempts.

Question 2

Why do online portals matter so much in customer identity programmes?

Accepted Answer

Portals are where customers experience whether identity design actually works. If access, recovery, and session handling are clumsy, users often read that as a lack of reliability or control. Strong portal design improves service confidence because it gives customers a clear, repeatable way to view and manage their own data.

Question 3

How can practitioners tell whether their team is reading deeply enough?

Accepted Answer

A useful test is whether the team can explain the mechanism, the assumptions, and the limits of a recommendation in its own words. If it cannot, the team is probably consuming information faster than it is understanding it.

Question 4

Why does Active Directory still matter to modern identity programmes?

Accepted Answer

Active Directory still matters because it often acts as the authoritative control plane for authentication, group-based authorisation, and privileged administration. When directory trust is centralised, its availability and integrity affect user access, service accounts, and recovery operations across the enterprise. Identity teams should treat it as core security infrastructure, not an ageing back-office service.

Question 5

What breaks when purchases and identity are not unified across channels?

Accepted Answer

When purchases and identity are not unified, users can buy value on one channel and fail to see it on another, which looks like fraud or product failure from their perspective. Support teams then inherit the inconsistency, and trust drops quickly. Unified identity is what keeps entitlements recognisable across devices.

Question 6

Why do customer identities need different controls from workforce identities?

Accepted Answer

Customer identities behave at much larger scale and under far less organisational context than workforce identities. That changes recovery, assurance, privacy, and user-experience requirements. A control that works for an employee can create abandonment or compliance issues for a customer, so the design must reflect the identity subject and the business outcome.

Question 7

Why do RAG systems need more than a single quality score?

Accepted Answer

A single score can hide whether the retriever failed, the generator hallucinated, or the answer was correct but poorly explained. Separate signals for grounding, relevance, and retrieval quality help teams identify the real problem and avoid fixing the wrong layer of the pipeline.

Question 8

How should security teams reduce risk from dormant and orphaned identities?

Accepted Answer

Start by identifying identities with no active owner, no recent use, or no business dependency, then revoke access in order of privilege and exposure. The goal is to remove accounts that can still authenticate but no longer serve a legitimate function. Lifecycle offboarding is the control that prevents these identities from becoming persistent attack paths.

Question 9

How should security teams handle dormant accounts without leaving downstream access behind?

Accepted Answer

Treat the dormant account as the starting point, not the finish line. Disable the identity, then trace linked application roles, tokens, delegated permissions, and service dependencies so access does not persist in systems the IdP cannot see. The goal is to remove entitlement residue, not just inactive usernames.

Question 10

Why do lifecycle models break down for non-human identities?

Accepted Answer

Because NHI lifecycles are driven by workload change, application retirement, vendor relationships, and key expiry rather than employee movement. If offboarding remains tied to HR events or quarterly reviews, credentials outlive their business purpose and remain usable long after they should have been revoked.

Question 11

What is the difference between storing secrets in docker-compose files and injecting them at runtime from a secrets manager?

Accepted Answer

Embedding secrets in compose files makes them part of source code and version history, which expands exposure and complicates rotation. Runtime injection keeps credentials out of the repository and allows centralized control, rotation policies, and access logging. The practical difference is whether the secret is treated as code artefact or as protected operational data.

Question 12

What breaks when encryption keys are not rotated and retired on schedule?

Accepted Answer

When keys are not rotated and retired on schedule, the exposure window stays open longer than policy allows. Compromised or overused keys can persist in applications, backups, and integrations, making breach impact harder to contain. Teams also lose confidence in key provenance, which weakens incident response, audit readiness, and the integrity of encrypted data over time.

Question 13

What breaks when user provisioning and de-provisioning are handled manually in IAM programmes?

Accepted Answer

Manual lifecycle management tends to create delays, inconsistent entitlements, and orphaned access that survives role changes or departures. Those errors undermine least privilege and make it difficult to prove access was removed on time. Automated provisioning and de-provisioning helps close those gaps, improves consistency, and reduces the chance of audit findings tied to stale access.

Question 14

How should security teams implement secrets detection and revocation across the SDLC?

Accepted Answer

Security teams should scan code, containers, and pipelines continuously, then revoke exposed credentials immediately before they can be abused. The best results come from automation at commit, pull request, and deployment stages, combined with clear ownership for remediation. Secrets handling works best when discovery, validation, and revocation are built into normal delivery rather than treated as a separate cleanup task.

Question 15

What breaks when certificate revocation is not enforced for external workload identities?

Accepted Answer

A revoked certificate can continue to authenticate if revocation checks are stale or incomplete, which turns a cut-off control into a paper exercise. The workload may keep assuming roles and reaching sensitive AWS resources long after operators believe access has ended. That is why revocation must be tested in the live access path, not just recorded in policy.

Question 16

How do security teams know if Roles Anywhere profiles are too permissive?

Accepted Answer

They are too permissive when the certificate can assume roles that reach data or services unrelated to the workload's documented function. Look for broad profile matching, shared trust anchors, and role assumptions that cannot be tied to a specific business service. If the access path cannot be explained in one sentence, it is probably too broad.

Question 17

What breaks when an Android app stores authentication tokens the wrong way?

Accepted Answer

Putting access tokens or refresh tokens in plain storage, logs, or crash reports turns a local credential into a reusable session for anyone who can read those artefacts. Access tokens should stay in memory because they are short lived. Refresh tokens should be encrypted at rest, rotated correctly, and cleared when sign out or revocation happens.

Question 18

How should security teams implement mobile sign-in when the app cannot safely store a client secret?

Accepted Answer

Use an OAuth style flow designed for public clients, with PKCE, a browser based sign-in surface, and a redirect URI that returns into the app. Keep only short lived access tokens in memory and store refresh tokens in encrypted, device backed storage. The app should never ship a secret, because anything inside an APK can be extracted.

Question 19

What breaks when organisations rely on certificate managers for post-quantum readiness?

Accepted Answer

They miss most of the cryptographic estate. Certificate managers usually track public TLS, but post-quantum readiness also depends on internal PKI, SSH keys, code signing, embedded algorithms, HSM-backed keys, and workload identity material. If those assets are invisible, the migration plan is built on partial data and the deadline will slip.

Question 20

Why do shrinking certificate lifetimes make manual renewal risky for enterprise workloads?

Accepted Answer

Shorter certificate lifetimes sharply increase the number of renewals teams must complete each year, which turns human tracking into an operational bottleneck. As certificates are used by services, load balancers, and workloads, missed renewals can cause outages and authentication failures. Automation becomes necessary because the renewal volume is too high for spreadsheets and ticket-driven processes to manage reliably.

Question 21

Why do organisations need MFA for cloud and customer-facing access even when passwords are already in place?

Accepted Answer

Passwords alone are vulnerable to phishing, reuse, replay, and credential theft. MFA adds an additional verification layer, so a stolen password does not automatically grant access. That matters most for cloud dashboards, customer portals, and transactional systems where one compromised account can expose sensitive data, disrupt services, or trigger regulatory consequences.

Question 22

What breaks when agents can call tools without shared runtime standards?

Accepted Answer

Without shared runtime standards, the same tool must be integrated and governed separately for each agent framework. That increases connector sprawl, makes policy enforcement uneven, and weakens auditability because each integration can log differently. In practice, teams lose a single view of agent behaviour, and security controls start to depend on implementation details instead of policy.

Question 23

What is the difference between a self-hosted private vault and a managed vault with customer-managed keys?

Accepted Answer

A self-hosted private vault gives the organisation direct control over the entire stack, including operations and audit readiness. A managed vault with customer-managed keys keeps key ownership with the customer while shifting infrastructure operation, monitoring, and independent assurance to the provider. The practical trade-off is less self-sovereignty in exchange for stronger continuous oversight.

Question 24

How should security teams decide whether a PAM vault needs HSM-backed protection or a fully managed vault service?

Accepted Answer

Start with the assurance level you need, not the component you want to buy. An HSM strengthens key protection, but it does not create independent audit, continuous monitoring, or operational depth. If you need self-sovereignty and can accept self-attested controls, HSM-backed self-hosting may be enough. If you need continuously verified assurance, a managed vault service is the stronger fit.

Question 25

What breaks when organisations treat a strong HSM as proof that the whole vault is secure?

Accepted Answer

The main failure is assuming a certified component guarantees system-level assurance. An HSM protects key material within its own boundary, but the vault still depends on host hardening, privilege controls, fault tolerance, and operational monitoring. If those layers are weak, a compromised server or admin path can still expose secrets even when the HSM itself is sound.

Question 26

Why do GenAI-driven social engineering attacks increase account takeover risk?

Accepted Answer

They compress the distance between persuasion and authorization. Attackers can research victims, generate tailored messages, and maintain a believable conversation until staff approve a reset or credential change. Once that happens, the attacker has a legitimate access path, which makes later fraud, persistence, or lateral movement much easier.

Question 27

How can security teams measure whether help desk identity assurance is working?

Accepted Answer

Look at the rate of resets denied because proofing failed, the number of recovery actions requiring callback validation, and the percentage of privileged requests that complete without manual discretion. If every urgent request can still be pushed through by conversation alone, the control is not working.

Question 28

What breaks when help desk recovery relies on voice or conversational trust?

Accepted Answer

Voice and conversational trust break because GenAI can imitate both with enough fidelity to defeat human intuition. If the process accepts plausibility as proof, an attacker only needs a convincing script or cloned voice to trigger a reset, re-enrol a device, or obtain access. Recovery must be treated as an authentication decision, not a support convenience.

Question 29

Who is accountable when a compromised local model runtime causes persistent AI agent compromise?

Accepted Answer

Accountability should sit with the teams that own the runtime, network exposure, and agent governance controls. Security, platform, and application owners must verify how the model is bound, who can modify templates, and whether browser-accessible attack paths exist. If those controls are missing, the organisation has not separated endpoint protection from runtime integrity.

Question 30

What breaks when teams do not discover hidden identities across cloud, servers, and Active Directory?

Accepted Answer

When hidden identities are not discovered, organisations lose control over who or what can authenticate, rotate, or be revoked. That leads to orphaned accounts, standing privilege, and inconsistent policy enforcement. The result is weaker incident response, slower cleanup after compromise, and a larger window for misuse.

Question 31

Who should be accountable for a human risk program when privacy and legal concerns are involved?

Accepted Answer

Security should sponsor the effort, but accountability must be shared across identity, privacy, legal, communications, business leaders, and managers. One program owner should hold decision rights, while privacy and legal define data boundaries and permitted use. This prevents the program from becoming covert surveillance and keeps interventions defensible.

Question 32

Why do self-built PAM vaults often fail to reach the same assurance level as institutional-grade vault services?

Accepted Answer

Self-built vaults can be strong, but they usually stop short because the hardest controls are operational, not technical. Independent audit, 24/7 monitoring, recurring assurance work, and trained split-knowledge staffing are expensive to sustain. A customer can buy hardware, but it is much harder to replicate a dedicated security organisation running those controls continuously across every tenant.

Question 33

Why does framework fragmentation increase risk in production agent environments?

Accepted Answer

Framework fragmentation creates separate auth paths, tool wiring, and audit gaps for each stack component. As teams add more agents, models, and tools, control becomes inconsistent and visibility drops. The risk is not any one framework. It is the accumulation of different governance models that makes it harder to prove who did what, when, and with which data.

Question 34

How should security teams govern agent interoperability across multiple frameworks and tool standards?

Accepted Answer

Security teams should put identity, access, audit, and guardrails in one control plane, rather than letting each framework govern itself. That means registering every agent centrally, enforcing role-based access to models and tools, and inspecting tool calls and responses at the gateway. Interoperability is only safe when governance stays consistent as frameworks, models, and connectors change.

Question 35

Why do data governance programs fail when teams begin with the framework instead of the problem?

Accepted Answer

They fail because different teams often mean different things by governance. One group may care about access, another about quality, and another about AI usage. If the program starts with tooling or policy templates, it can solve the wrong issue, spread accountability thin, and leave the actual business risk untouched.

Question 36

Why do unmanaged service accounts and AI-related credentials create so much risk in modern environments?

Accepted Answer

Unmanaged non-human identities expand the attack surface because they often outnumber human accounts, persist longer, and escape normal governance. When credentials are scattered across cloud, servers, and code, attackers can find dormant access paths, reuse secrets, and move laterally without triggering the controls built for user accounts.

Question 37

Who should own response when a support account moves confidential archives to a personal channel?

Accepted Answer

Identity, endpoint, and insider-risk teams should share ownership because the event spans access, device trust, and data movement. The response should focus on containment of the account, review of the destination channel, and validation of whether the session was deliberate or compromised.

Question 38

How should security teams detect exfiltration when every file transfer is individually allowed?

Accepted Answer

Correlate the full sequence, not the per-file verdict. A burst of downloads from a sensitive SaaS followed by uploads to a new personal destination is the exfiltration signal, especially when the destination is first-time for that account and the transfer volume exceeds its normal baseline.

Question 39

Why do authorised users still create serious data-loss risk in managed environments?

Accepted Answer

Because authorisation to access data is not the same as authorisation to remove it in bulk. When a legitimate account can gather a coherent record set and send it to an external destination, the organisation has an identity and sequence problem, not just a malware problem.

Question 40

How should security teams defend against authorization phishing when passkeys and MFA are already in place?

Accepted Answer

Teams should treat passkeys and MFA as authentication controls, not complete protection against post-login abuse. Authorization phishing targets OAuth consent, device code, and token exchange flows after login succeeds. The right response is browser-layer detection, device code flow restrictions where feasible, scoped conditional access, and validation through live simulations. Security teams should assume authentication strength alone does not close the…

Question 41

What fails when teams rely on point-in-time vulnerability scans for internet-facing systems?

Accepted Answer

They get a snapshot of historical exposure, not proof of current exploitability. Attackers use the gap between scans to target systems that have already changed, especially exposed collaboration and edge services. Continuous validation is needed when risk depends on whether a flaw is reachable right now, not whether it appeared in last week’s report.

Question 42

When should organisations prioritise CTEM over faster patch cycles?

Accepted Answer

CTEM should move up the list whenever exposed assets can be targeted faster than your assessment cadence, especially on internet-facing services with unauthenticated attack paths. Faster patching still matters, but it does not answer whether a given asset is exploitable today. Validation must come before confidence.

Question 43

How do security teams detect when an approved agent has drifted outside its intended role?

Accepted Answer

Teams need policies that compare the agent’s recorded intent with its current capabilities and runtime behavior. If an agent approved for one purpose later acquires an irreversible tool, broader data access, or outbound paths it was not meant to have, the drift should trigger a posture issue and, where possible, runtime enforcement.

Question 44

What breaks when an agent inventory is managed only with manual review and spreadsheets?

Accepted Answer

Manual review breaks down once many teams build agents independently and each agent connects to multiple tools or MCP servers. Labels become stale quickly, reviewers lose a reliable view of who can do what, and security teams cannot scale approval, risk detection, or drift monitoring across the full inventory.

Question 45

Why do AI agents need durable state outside the context window to function safely as collaborators?

Accepted Answer

AI agents need durable state because their context window is temporary and can lose decisions, constraints, and prior work as tasks continue. External memory lets teams preserve what changed, what was decided, and what remains open. That shared state supports handoffs, auditability, and recovery when an agent resumes work or a run fails.

Question 46

What is the difference between identity controls and guardrails in AI agent governance?

Accepted Answer

Identity controls answer who or what may act. Guardrails answer what that action may contain. In practice, identity governs trust boundaries, while guardrails inspect the actual payload moving through those boundaries. Both are needed for agentic systems because a fully authorized agent can still be tricked into sending unsafe content, leaking secrets, or following injected instructions.

Question 47

Why do AI agents need guardrails even when identity and access control are already in place?

Accepted Answer

Identity and access control decide whether an agent is allowed to call a model or tool. Guardrails decide what that call is allowed to contain. That matters because a valid credential can still carry an injected instruction, a destructive SQL argument, or a leaked secret. The risk is not authorization alone, but unsafe content becoming unsafe action.

Question 48

What breaks when CSPM and DSPM are used alone for AI security?

Accepted Answer

They can identify misconfigurations and sensitive data, but they do not tell you which agent can act on that data or which tool chain can reach it. Without identity context, teams may know where the data is, but not who or what can use it.

Question 49

Why does identity context matter more when AI agents enter the enterprise?

Accepted Answer

Because agentic systems can scale access decisions, tool use, and data reach much faster than manual governance can track. Without context such as owner, scope, and blast radius, teams cannot tell whether an agent is operating within approved limits or quietly expanding exposure across systems.

Question 50

Should organisations prioritise privacy-preserving verification over biometric proofing?

Accepted Answer

Yes, when the business requirement can be met without biometric capture. Privacy-preserving verification reduces the amount of sensitive data in scope, lowers the impact of breaches and helps preserve trust in communities that rely on anonymity or pseudonymity.

Question 51

What breaks when age verification is outsourced to a third party?

Accepted Answer

Outsourcing does not remove accountability. It creates a new dependency on the vendor’s support, storage and deletion controls, and any failure there still affects the platform’s users and reputation. The platform must govern the processor lifecycle, not just the policy outcome.

Question 52

Why do AI deepfakes increase fraud risk even when people are trained to spot them?

Accepted Answer

Training helps, but it cannot reliably overcome synthetic voice, video, and writing that mimic familiar people under pressure. Deepfakes exploit urgency and authority, which are stronger than visual or verbal cues in many business settings. The risk rises when organisations let a convincing message trigger an irreversible outcome without a second verification step.

Question 53

What breaks when executive requests can bypass normal verification?

Accepted Answer

The control breaks at the point where trust replaces evidence. Once staff believe a senior request can override process, attackers only need to imitate authority well enough to create urgency. That turns approvals into a social decision instead of a governed one, which is why wire transfers, payment changes, and access exceptions are the most exposed.

Question 54

How do security teams know if impersonation detection is actually working?

Accepted Answer

It is working if high-risk requests are consistently challenged, blocked, or escalated when the claimant cannot satisfy stronger proofing controls. Teams should look for reduced success rates on password resets, MFA bypass attempts, and account recovery requests that originate from unmanaged or suspicious channels. If support agents still rely on familiar questions alone, the control is not effective.

Question 55

What breaks when organizations rely on knowledge-based verification for AI-powered fraud?

Accepted Answer

Knowledge-based verification breaks because the answers are often guessable, breached, or publicly available, especially when attackers use AI to sound convincing in real time. That makes help desks and call centers easy targets for impersonation. The control fails at the moment of trust, not after the fraud is complete, so organisations need stronger proofing methods before sensitive access is granted.

Question 56

Why do deepfake attacks change fraud verification risk?

Accepted Answer

Deepfakes change the risk because they attack the assumptions behind liveness detection. A control built to spot a fake photo or simple spoof can still accept synthetic video, cloned audio, or injected streams. That means fraud teams must shift from presence checks to provenance and trust checks.