AI impersonation is exposing gaps in enterprise identity controls

At a glance

What this is: This article argues that generative AI has made impersonation attacks more convincing and more scalable, turning identity signals themselves into a target.

Why it matters: It matters because IAM, PAM, and NHI programmes now need to verify behaviour and session context, not just initial identity proofing or MFA prompts.

By the numbers:

• CrowdStrike’s 2025 Global Threat Report revealed a 442% spike in AI-powered voice phishing attacks in just six months.

Context

AI impersonation is a governance problem as much as a fraud problem. The article describes attackers using synthetic voice, text, and video to pass as employees, vendors, or executives, which means the control failure starts when identity signals are trusted too early and too absolutely.

For IAM teams, the important shift is not the presence of deepfakes alone but the way they interact with authentication, authorisation, and user verification. Once an attacker can convincingly imitate a trusted person, the programme has to treat session behaviour, token use, and approval paths as part of identity assurance.

Key questions

Q: How should security teams defend against AI-powered impersonation attacks?

A: Security teams should combine strong identity verification with continuous monitoring and tight authorization limits. Use out-of-band confirmation for high-risk actions, shorten session lifetimes, revoke tokens quickly, and log every sensitive approval. The best defence is not a stronger login alone, but a control stack that limits how far a convincing impersonation can travel once trust is granted.

Q: Why do passwords and one-time MFA struggle against deepfake impersonation?

A: Passwords and one-time MFA verify a moment, not the legitimacy of every request that follows. Deepfake voice or text can manipulate help desks, vendors, and employees into approving resets or transfers, so the attacker wins by abusing trust after initial verification rather than by breaking the authenticator itself.

Q: What are the warning signs that an impersonation attack is succeeding?

A: Look for unusual device changes, impossible travel, repeated login attempts, sudden approval requests, and requests that bypass normal verification channels. A pattern of social pressure combined with identity events that do not match the user’s normal workflow is a strong indicator that impersonation is in progress.

Q: What should teams do when an executive or vendor request looks suspicious?

A: Pause the request, verify it through a separate trusted channel, and require step-up approval before any payment, access change, or account recovery. The goal is to stop a convincing impersonation from becoming a privileged business action before the request completes.

Technical breakdown

How AI impersonation defeats one-time verification

AI-crafted phishing, cloned voices, and manipulated video compress the attacker effort required to look legitimate. That matters because many identity controls still treat the first successful login, approval, or support interaction as a strong trust event. Once that trust is established, the attacker can reuse the same social signal across email, voice, chat, and video to push victims toward credential entry, payment approval, or reset requests. The technical issue is not only better-forged media. It is the way these forgeries are used to trigger valid identity workflows that were designed around human-paced verification and stable trust assumptions.

Practical implication: move verification from a single challenge-response moment to repeated checks across the session and request lifecycle.

Why replay attacks and token abuse matter in identity fraud

The article notes automated credential testing, replayed SAML assertions, and API probing as part of the attacker playbook. These tactics matter because modern identity systems often separate proof of identity from proof of intent, then allow tokens or assertions to carry trust across multiple actions. If a token can be replayed, or if a support workflow can be manipulated into issuing a fresh credential, impersonation becomes a direct path to authorisation. This is especially dangerous in environments that rely on federated sign-in and broad downstream access after initial authentication.

Practical implication: treat token replay resistance, short session lifetime, and revocation speed as core fraud controls, not just technical hardening.

Where malicious AI agents fit into the attack model

The article also points to rogue AI agents that blend into normal workflows while exfiltrating data. That is a different problem from simple automation. Here, the attacker uses software that can behave like a legitimate operational actor, which makes detection harder because the activity may resemble approved system behaviour. For identity governance, this widens the scope from human impersonation to machine-like impersonation, where the real question is whether the workflow is allowed to initiate actions that look trustworthy by default. The result is a larger trust boundary around every delegated action.

Practical implication: inventory and govern delegated workflow identities separately from human users and do not assume automation is inherently benign.

Threat narrative

Attacker objective: The attacker’s objective is to obtain trusted access or approval that can be converted into money movement, account compromise, or sensitive data theft.

1. Entry begins with AI-generated phishing, vishing, or executive impersonation that convinces a target to engage with a fraudulent request.
2. Credential access follows when the target reveals credentials, approves a payment, or accepts a manipulated support or login prompt.
3. Escalation occurs when the attacker reuses trusted identity signals, replays assertions, or tests stolen tokens against enterprise systems and APIs.
4. Impact is realised through fraudulent transfers, data exposure, or broader account compromise that erodes trust in identity-based controls.

Breaches seen in the wild

• Entra ID actor token flaw (CVE-2025-55241): Hidden Actor tokens plus an Azure AD Graph validation flaw could have let attackers become Global Admin in any Entra ID tenant (CVE-2025-55241).
• Gemini CLI prompt injection flaw 2025: Tracebit showed a poisoned README could make Gemini CLI run hidden commands and exfiltrate developer secrets; Google fixed it in 0.1.14.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 150+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

AI impersonation turns identity assurance into a continuous verification problem: The article shows that the trust event is no longer the login screen but every approval, support exchange, and follow-on action that can be socially engineered. That means identity programmes built around single verification moments are operating on a broken assumption. Practitioners should treat identity as a live control surface, not a one-time checkpoint.

Session control is now part of fraud prevention, not just access management: When attackers can replay assertions, manipulate support workflows, or ride a valid session after impersonation, the failure is not authentication alone. The deeper gap is weak binding between identity proof, intent, and session continuity. The implication is that identity security teams must evaluate whether their trust model survives handoff from initial sign-in to downstream action.

Continuous behaviour monitoring is becoming an identity control, not an add-on: The article’s emphasis on anomaly detection, device signals, and session logging reflects a category shift. Static rules fail when the attacker can vary the content, medium, and timing of impersonation. The practical conclusion is that identity governance now has to include behavioural evidence, not just account status and MFA success.

Named concept: identity signal spoofing debt: This article exposes the accumulated risk created when organisations keep trusting voice, text, and video as if they were stable identity proofs. That assumption works only when attackers cannot cheaply synthesize trusted signals at scale. Once they can, the debt shows up as fragile approval paths, weak help-desk controls, and overconfident authentication states.

Machine impersonation expands the NHI boundary: The article is not only about humans being spoofed. It also shows rogue AI agents blending into operational workflows, which means delegated identities and automated actors need the same governance seriousness as human accounts. The implication is clear: if a system can speak and act credibly on behalf of someone or something else, it belongs in identity governance scope.

From our research library:

• Veriff's 2026 Identity Fraud Report documented a 300X increase in digitally presented media that was either entirely AI-generated or otherwise altered.

What this signals

Identity assurance now has to include the medium of the request, not just the account behind it: Voice, text, and video can all be forged well enough to pass as trusted identity signals. That means verification controls need to evaluate how a request arrives and whether its delivery path matches the expected actor, not simply whether the actor name is known.

Continuous session scrutiny is becoming the real control boundary: Once an impersonator has passed a first check, the remaining risk sits in the session, token, and approval path. Organisations that still rely on static trust decisions will miss the moment when a legitimate-looking interaction turns into a fraudulent one.

For practitioners

• Strengthen SSO trust boundaries Centralise authentication through SAML or OIDC, then pair sign-in with device and anomaly signals so a single successful login does not create open-ended trust.
• Shorten session trust windows Apply strict session lifetimes, fast token revocation, and audit-grade logging so a stolen or replayed credential cannot remain useful long after first use.
• Replace static verification with continuous checks Verify sensitive actions through behaviour, context, and approval path changes instead of relying on one-time identity proofing or a single MFA event.
• Tighten privileged approval paths Require separate verification for payment changes, account resets, and vendor requests so impersonation cannot convert a social success into privileged action.
• Inventory delegated and automated identities Track AI-driven workflows, bots, and service identities as governed actors so impersonation controls extend beyond human users.

Key takeaways

• AI impersonation creates fraud and access risk because attackers can now imitate trusted people well enough to trigger normal identity workflows.
• The article cites a 442% spike in AI-powered voice phishing attacks in just six months, showing that the threat is scaling quickly.
• Defences need to shift from one-time verification to continuous monitoring of session behaviour, token use, and approval intent.

Key terms

• AI-powered impersonation: The use of generative or adaptive AI techniques to mimic legitimate users, documents, or behaviours. In identity operations, this raises the quality and speed of deception, making static rules and isolated checks less reliable as primary controls.
• Continuous Verification: A Zero Trust practice that re-evaluates trust during the session instead of relying on a single successful login. The control is stronger when context signals are available in real time and when the identity programme can act on those signals without creating excessive exceptions.
• Session trust window: A session trust window is the period during which an authenticated identity is allowed to keep acting without renewed scrutiny. AI impersonation shortens the safe window dramatically because attackers can abuse a valid session or approval path after the first verification step has already succeeded.
• Identity Spoofing: Identity spoofing is the use of false or stolen identity signals to make an AI system or its surrounding controls treat an untrusted actor as legitimate. In AI environments, it often combines with token abuse, weak service-account design, or poor trust boundaries.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity lifecycle 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 or identity governance programme, it is worth exploring.