NHI & Agentic AI Identity Security FAQ

Question 1

What is the difference between human identities and non-human identities in enterprise security?

Accepted Answer

Human identities represent people who authenticate to work systems, while non-human identities represent applications, services, scripts, and workloads that authenticate on their own. Both need authentication, authorization, and management, but the technology stack differs because machines often use keys, tokens, certificates, or service accounts instead of passwords. That distinction matters for governance, monitoring, and lifecycle control.

Question 2

What are the signs that an MFA approach is creating more operational burden than security value?

Accepted Answer

Common warning signs include repeated account lockouts, delayed OTP delivery, frequent password resets, help desk tickets about login problems, and users struggling with app switching or device dependency. If authentication steps increase support demand and slow daily work without materially improving resistance to phishing, the organisation is paying for complexity rather than stronger assurance.

Question 3

What is the difference between federated identity and decentralized identity for enterprise access?

Accepted Answer

Federated identity uses a trusted identity provider to authenticate users once and then grant access to multiple services through SSO. Decentralized identity shifts credential control to the user, often through verifiable credentials and digital wallets. The first optimises centralized oversight and ease of administration, while the second prioritises user control and reduced reliance on a central gatekeeper.

Question 4

What is the difference between post-quantum cryptography and quantum-generated randomness in secure identity systems?

Accepted Answer

Post-quantum cryptography protects data and authentication against future attacks from quantum computers, while quantum-generated randomness strengthens the unpredictability used to create keys and other security material. They solve different problems. One hardens the cryptographic algorithms themselves, and the other improves the quality of the entropy feeding those algorithms, which is critical for trusted identity and key provisioning.

Question 5

How do you know if a confirmation flow is actually phishing resistant?

Accepted Answer

Look for three things: the proof is bound to the right session, it requires an enrolled device or equivalent cryptographic proof, and it expires quickly after use. If any of those are missing, the flow may still be convenient, but it is not materially resistant to relay or forwarding attacks.

Question 6

Who should approve fallback access when device proof is unavailable?

Accepted Answer

Fallback access should be approved only by a separate control owner, not the same agent handling the call. If the organisation must allow exceptions, they should be slower, fully logged, and reviewed as privileged decisions because fallback paths are where attackers will concentrate pressure.

Question 7

How should teams prevent oversharing in identity verification workflows?

Accepted Answer

Start by defining a minimum disclosure contract for each workflow and require the presenter to preview the exact fields before release. Then enforce mandatory provenance markers, so verified and self-asserted attributes are not treated as equal. Finally, make sensitive fields step-up gated and keep the request single-use and time-limited.

Question 8

What breaks when passwords and secrets are spread across personal tools instead of a central control plane?

Accepted Answer

When secrets are spread across personal tools, teams lose central visibility, consistent permissioning, and reliable offboarding. Orphaned access can remain after employee exit, password health becomes harder to enforce, and manual sharing through email or Slack increases the chance of leakage. The operational result is slower remediation, weaker governance, and more blind spots.

Question 9

Why do password resets and insecure sharing practices create so much operational risk?

Accepted Answer

Password resets consume time for both the locked-out employee and IT support, and insecure workarounds like email or messaging expose credentials outside controlled access paths. A secure password manager reduces both problems by keeping credentials in one encrypted vault and enabling controlled sharing. That improves productivity while lowering the chance that business logins are reused, exposed, or handled inconsistently.

Question 10

What mistakes do teams get wrong when using secure notes in a password vault?

Accepted Answer

A common mistake is creating a note with only a title and leaving the sensitive content outside the note body. Another is treating the vault as a dumping ground without adding enough structure, such as custom fields, folders, or renewal details. Teams also overlook access hygiene, even though the note may contain information that deserves the same discipline as passwords.

Question 11

How should security teams decide when to use secure notes instead of password entries in a vault?

Accepted Answer

Security teams should use secure notes when the information is sensitive but does not fit a username, password, card, or identity record. That includes license keys, bank details, insurance data, stock information, and other unstructured records that still require encryption and controlled access. The goal is to keep sensitive operational data in one protected vault, not scattered across documents or…

Question 12

What is the difference between saving a login manually before account creation and using the autosave banner afterward?

Accepted Answer

Manual saving works best when teams want to generate credentials upfront and use autofill during signup. The autosave banner is better when the account has already been created and the login needs to be captured afterward. Both methods end with a stored credential in the vault, but they differ in timing, user workflow, and how much editing is needed before…

Question 13

What should organisations do when a new account is created before the credential already exists in the vault?

Accepted Answer

Use the autosave banner to capture the login immediately after signup, then review the saved item before relying on it. If the banner offers an edit path, confirm the username, password, destination folder, and any organisation collection before saving. This preserves vault accuracy while still letting users complete registration without manual reentry later.

Question 14

How should security teams create and save new login credentials so users can sign up quickly without weakening password hygiene?

Accepted Answer

The best pattern is to generate the username and password before finishing account creation, save the item in the vault, and then use autofill during signup. This reduces reuse, keeps credentials centrally stored, and avoids copy paste mistakes. Teams should also make sure generator settings match the site’s password rules so the first saved credential is both strong and usable.

Question 15

What is the difference between phishing resistant authentication and OTP-based MFA in an adversary-in-the-middle attack?

Accepted Answer

OTP-based MFA can still be relayed because the code is reusable for the brief window it is valid, so a proxy attacker can pass it through to the legitimate service. Phishing resistant authentication binds the proof to the real origin and the device or key, which prevents simple replay through a fake login page.

Question 16

What is the difference between runtime authorization and traditional vault-based credential checkout?

Accepted Answer

Traditional vault checkout stores credentials until someone or something retrieves them, which can leave standing access in place longer than necessary. Runtime authorization changes the model by evaluating each request in real time, issuing a short-lived credential only for the approved action, and revoking it immediately afterward. The difference is not storage alone, but when trust is granted and how…

Question 17

What are the signs that an authentication journey is relying too heavily on SMS OTP?

Accepted Answer

A journey is overdependent on SMS OTP when users regularly wait on messages, abandon flows for manual entry, or need repeated retries because delivery fails. The control is also misaligned if the organisation lacks alternatives for restricted authenticator methods. Those signals show the experience and assurance model both depend on a fragile, single verification path.

Question 18

What is the difference between number verification and SMS OTP in authentication?

Accepted Answer

Number verification confirms possession of a mobile number without asking the customer to receive and enter a one-time passcode. SMS OTP still sends a secret that must be typed back into the journey. The difference matters because number verification removes the code-sharing step that phishing attacks target, while SMS OTP keeps that exposure in place.

Question 19

What is the difference between static MFA policies and adaptive access policies in identity security?

Accepted Answer

Static MFA policies apply the same authentication requirement to everyone or to broad rule groups. Adaptive access policies evaluate risk continuously using behavioral, device, and contextual signals, then decide whether to allow access, require step-up authentication, or block the request. Adaptive policies are better suited to fast-changing AI-driven attacks.

Question 20

Why does forwarding a bearer token unchanged create both a security risk and an audit problem in agent workflows?

Accepted Answer

Because the downstream service logs whatever identity arrives, not the original user or intent behind the request. If the token was not issued for the receiving service, the log inherits a misleading identity and the system may act with broader authority than intended. That weakens per-hop authorization and makes later investigation depend on inference instead of evidence.

Question 21

What is the difference between login with SSO and SSO with customer managed encryption?

Accepted Answer

Login with SSO uses the identity provider for authentication while the user still enters a master password to decrypt the vault. Customer managed encryption removes the master password and retrieves the vault key from a self-hosted key server. The first keeps decryption anchored to the user, while the second shifts key custody to the customer’s own infrastructure.

Question 22

How should IT teams centralise unmanaged company passwords and secrets without creating more user friction?

Accepted Answer

IT teams should move credentials into a single encrypted vault with browser and mobile access, then layer in controlled sharing, automatic sign in, and role based administration. The goal is to replace scattered storage in spreadsheets, chat threads, and personal tools with one source of truth that improves visibility, reduces exposure, and preserves day to day usability.

Question 23

How should organisations secure AI workflows when public inputs can reach privileged steps?

Accepted Answer

Treat workflow transitions as trust boundaries, not internal plumbing. Every step that can receive a public request should verify identity, enforce least privilege, and separate requester permissions from the permissions used by the AI task itself. Replace static administrative keys where possible, log requests and outputs, and monitor which workflows can touch sensitive data. If a public interface exists, assume…

Question 24

Why do security teams need defence in depth across password manager software, infrastructure, and processes?

Accepted Answer

Password management cannot be secured by a single control layer because risk exists in the application, the hosting environment, and the operational process. A defence in depth model reduces reliance on any one safeguard and improves resilience if one layer fails. That approach is especially important for systems storing logins, secure notes, and other sensitive credentials.

Question 25

Why do passwordless SSO designs change the risk profile for vault access and key recovery?

Accepted Answer

Passwordless SSO reduces password fatigue and forgotten credential lockouts, but it shifts trust toward device assurance or customer managed key infrastructure. That changes the failure domain. If the device is compromised or the key server is weakly governed, the vault can still be exposed. The security model depends on strong controls around trusted endpoints and encryption key custody.

Question 26

How should security teams implement SSO with trusted devices without weakening account recovery controls?

Accepted Answer

Security teams should treat SSO with trusted devices as an access convenience layer, not a replacement for account governance. Require clear device approval workflows, separate admin approval paths, and documented recovery options for users who never created a Bitwarden password. The key control is balancing passwordless access on approved devices with a recoverable account model and tight device trust management.

Question 27

Why do cloud native applications benefit from a relationship based authorization model?

Accepted Answer

Cloud native environments often involve many services, identities, and changing resource relationships. A relationship based model helps express who can access what in a way that matches those moving parts. That reduces brittle policy sprawl and makes authorization easier to reason about across microservices, containers, and platform components.

Question 28

How should security teams implement Microsoft Entra ID SSO with SAML in a multi-tenant application without creating inconsistent access policies?

Accepted Answer

Security teams should centralize identity policy in the IdP, then map SAML assertions into the application in a controlled way. The practical goal is consistent authentication, predictable access handling, and easier tenant administration. Teams should also validate setup in the admin portal, test enrollment flows, and confirm that role and group mappings preserve least privilege across tenants.

Question 29

What are the signs that a TIBER-EU programme is losing effectiveness over time?

Accepted Answer

A TIBER-EU programme is losing effectiveness when remediation lessons fade, new configurations reopen known paths, or teams can only explain resilience right after an assessment. If control checks are green but attack paths still exist, or if annual testing is the only evidence available, the organisation is relying on stale assurance rather than current proof.

Question 30

What are the signs that AI agent deployment is getting ahead of endpoint governance?

Accepted Answer

A common sign is tool adoption spreading across marketing, engineering, and sales with different update cadences and no formal IT review. Another is that agents are launched under whatever account is convenient, including admin accounts, so the real privilege boundary depends on the endpoint rather than policy. That creates inconsistent exposure and makes risk hard to track centrally.

Question 31

Who should own mule-risk controls when payments, fraud, and compliance teams all touch the same flow?

Accepted Answer

Ownership should sit with the team accountable for fraud prevention strategy, but execution must be shared. Payments teams control rail design, fraud teams tune scoring and detection, and compliance teams manage regulatory expectations and case handling. The practical answer is a single operating model with clear decision rights, shared data, and measurable escalation paths.

Question 32

What happens after the first PQC migration wave if teams do not build continuous crypto agility?

Accepted Answer

Risk returns once standards evolve, systems are added, and libraries change. A one-time cleanup may remove today’s exposure, but it does not keep the environment ready for tomorrow’s cryptographic shifts. Teams need continuous discovery and adaptation so new dependencies, new algorithms, and new weak points are identified before they become another migration project.

Question 33

What breaks when a PQC migration plan cannot handle blockers and non-migratable systems?

Accepted Answer

A migration plan breaks when it assumes every asset can be swapped on a clean schedule. Real environments include applications that do not support quantum-safe cryptography yet, vendors that are not ready, libraries that need code changes, and systems that may need replacement. Without a plan for exceptions, remediation stalls and backlog risk accumulates.

Question 34

How should security teams evaluate PQC readiness beyond a simple inventory of cryptography?

Accepted Answer

Teams should assess whether the tooling can discover cryptography across the environment and also explain what that cryptography supports. A useful PQC program needs relationships, not just objects. Inventory alone is incomplete if it cannot map certificates, libraries, algorithms, applications, business services, and ownership well enough to support prioritisation and migration planning.

Question 35

Why do siloed tools create more risk when organisations manage both human and non-human identities?

Accepted Answer

Siloed tools create risk because they break the link between identity, access, and governance across humans and machines. When access, security, and compliance are handled separately, teams lose visibility into what an agent can reach, who approved it, and whether its runtime actions still match intent. That fragmentation increases the chance of over-privileged, unmanaged automation.

Question 36

What are the signs that bonus abuse controls are not keeping up with fraud rings?

Accepted Answer

A common warning sign is when legitimate approval rates improve while bonus abuse block rates stall, which suggests the detection model is no longer adapting to new patterns. Another signal is clustering by geography, device type, payment method, or registration time. Those patterns often indicate coordinated activity rather than isolated misuse and should trigger re-tuning.

Question 37

Why does stronger liveness detection often create a trade-off between fraud prevention and user conversion?

Accepted Answer

Stronger liveness controls reduce spoofing by making it harder for photos, masks, replays, and deepfakes to pass. The trade-off is friction. Additional prompts, stricter models, and tighter thresholds can reject legitimate users in poor lighting, on older devices, or with brief capture errors. Teams must balance fraud loss against abandonment and support cost.

Question 38

What are the signs that a phishing campaign is using a custom-built reverse proxy rather than a public toolkit?

Accepted Answer

Common signs include self hosted bot checks, pixel accurate login flows, custom session handling, live field capture, and infrastructure that changes per target. In this case, certificate transparency logs exposed many related subdomains, and some looked like internal test traffic. Those operational traces often reveal more than the phishing page itself.

Question 39

Why do adversary-in-the-middle phishing kits remain effective even when users enter a one-time passcode?

Accepted Answer

They work because the attacker is not trying to crack the code. Instead, the phishing site relays the live username, password, and OTP to the real service as the victim submits them, then captures the resulting session. That makes MFA weaker when the factor is phishable, since the attacker can authenticate as the user in real time.

Question 40

How should security teams defend against adversary-in-the-middle phishing that relays credentials and one-time codes in real time?

Accepted Answer

The strongest defense is phishing resistant authentication, especially passkeys or hardware security keys, because the cryptographic challenge is bound to the real origin and cannot be replayed through a lookalike domain. Teams should also assume credentials and session cookies can be stolen from convincing reverse proxy kits, so monitoring, user awareness, and rapid takedown processes remain necessary.

Question 41

Why do compromised AI tools and OAuth grants create such a fast-moving identity risk?

Accepted Answer

Compromised AI tools become dangerous because they often sit on trusted sessions and broad OAuth permissions. Once attackers steal tokens or abuse granted scopes, they can pivot into email, drive, code, or internal systems before traditional controls notice. The risk is amplified when the tool can act like the user, because the activity looks legitimate until behaviour diverges.

Question 42

Why do generative AI and deepfakes increase account takeover risk for identity systems that rely on static checks?

Accepted Answer

Generative AI lowers the effort required to imitate legitimate users, create convincing synthetic identities, and automate probing at scale. Static checks tend to look for fixed patterns, so they miss fast-changing attack behavior. When identity controls do not incorporate live context and behavioral signals, attackers can blend in long enough to gain access.

Question 43

What happens when an AI agent is launched without scoped elevation but needs one protected installation step?

Accepted Answer

If the tool genuinely needs elevated rights for a one-time install and no scoped elevation exists, teams usually see workflow friction and more support tickets. The better pattern is to keep the account at standard user by default and grant elevation only for the specific installer or action. That limits the blast radius while preserving a workable install path.

Question 44

How should security teams reduce the risk of AI agents inheriting local admin rights on endpoints?

Accepted Answer

Security teams should treat AI agents like any other process that inherits the launching account’s token. The safest default is standard user access, not local admin, because the agent and any helper processes gain the same rights as the parent account. If elevation is needed, scope it narrowly to a specific installer or task, then return the account to standard…

Question 45

How should security teams govern AI agents and scripts as part of the identity perimeter?

Accepted Answer

Security teams should treat AI agents, custom scripts, and autonomous workflows as identities that need assignment, policy, and audit controls. The practical goal is to connect each digital worker directly to enterprise systems, tie activity to a known user or service owner, and restrict permissions to the minimum required. That reduces shadow automation and makes machine-speed actions traceable.

Question 46

What happens when an AI agent is blocked during a task?

Accepted Answer

The best response is not to end the session with a generic error. The agent should receive the reason for the denial, so it can adapt and continue the task within policy. That approach preserves productivity while preventing sensitive data leakage, destructive commands, and unauthorized transfers outside approved resources.

Question 47

What is the difference between account ownership and action-based identity governance for AI agents?

Accepted Answer

Account ownership asks who controls the identity on paper. Action-based governance asks what the identity is doing, why it needs access, and whether that activity is appropriate right now. For AI agents and other machine identities, the second model is more useful because risk changes with context, runtime behaviour, and the specific systems being touched.

Question 48

What happens when an AI agent is allowed to cross systems without clear delegated authority?

Accepted Answer

When an AI agent crosses systems without clear delegated authority, its permissions can expand beyond the original task, and the receiving system may not know what the agent is allowed to do. That creates gaps in enforcement, auditability, and accountability, especially when the workflow involves clinical decisions, privacy-sensitive data, or time-sensitive care coordination.

Question 49

Why do AI agents need identity context when tool calls are blocked or transformed?

Accepted Answer

Because a risky tool call is only meaningful when it can be tied back to a verified actor and an approved policy. Without identity context, teams see alerts but cannot tell which agent acted, on whose behalf, or whether the access was legitimate. Identity context turns isolated detections into actionable events and supports precise revocation without disrupting other agent connections.

Question 50

Why do KYC checks alone fail to stop multi-accounting in iGaming?

Accepted Answer

KYC verifies that a user appears to be who they claim at a single point in time, but it does not tell you whether the same person has already claimed the same promotion under another identity. Multi-accounting exploits that gap. Effective controls need journey-level correlation across device, payment, and behavior signals to reveal linked accounts.

Question 51

What is the difference between passive and active liveness checks in selfie verification?

Accepted Answer

Active liveness asks the user to blink, smile, or turn their head, then verifies the response in real time. Passive liveness works in the background from a still image or short video, using motion, texture, and depth cues without extra user action. Active checks add friction, while passive checks usually improve conversion but depend heavily on model quality.

Question 52

What are the signs that selfie verification is being used too narrowly?

Accepted Answer

A narrow deployment shows up when teams treat a selfie match as proof of identity on its own. Warning signs include forged documents slipping through, repeated fraud on approved accounts, weak device or network correlation, and no follow up monitoring after approval. If only the face check is measured, the surrounding fraud path is usually under controlled.

Question 53

What do fraud teams get wrong when they rely on traditional rules to catch mule activity?

Accepted Answer

The common mistake is assuming static rules can keep up with coordinated scam operations. Traditional monitoring often detects only obvious patterns after the money has moved. Mule activity is dynamic, cross-rail, and designed to stay below thresholds, so teams need intelligence at onboarding, screening, and investigation, not only alerts after settlement.

Question 54

Why are mule networks so effective across P2P, ACH, wire, remittance, and crypto payments?

Accepted Answer

Mule networks work because they act as the connective layer between rails. They move stolen funds quickly, split flows across multiple accounts, and exploit the speed gap between payment execution and traditional monitoring. That makes them difficult to see with isolated rules, especially when each rail is reviewed separately instead of as one laundering path.

Question 55

How should fraud teams shift from post-transaction review to pre-transaction risk scoring on instant payment rails?

Accepted Answer

Fraud teams should move risk decisions earlier, before funds leave the customer. On instant rails, post-transaction review is usually too late to stop loss. Pre-transaction scoring should combine onboarding data, payment context, mule indicators, and case history so high-risk transfers can be slowed, stepped up, or blocked before settlement occurs.

Question 56

How should supply chain teams verify drivers and carriers before releasing a shipment?

Accepted Answer

Teams should verify both the person and the business before handoff. That means checking the driver’s identity document, face match, and liveness, then confirming the licence, carrier authority, and the shipment assignment against trusted records. At pickup, the driver, vehicle, carrier, location, and time should all match the approved record before custody changes.