AI-enabled fraud is breaking identity assumptions in consumer auth

At a glance

What this is: This is an analysis of how AI is weakening knowledge-based authentication and why possession-based controls are being positioned as the more durable trust model.

Why it matters: IAM, fraud, and identity teams need to reassess authentication design because AI now scales credential theft, social engineering, and synthetic identity attacks faster than human-mediated controls can absorb.

By the numbers:

• AI-powered phishing relay attacks automate the interception in real time, and the entire flow takes under 90 seconds.
• Phishing volumes increased 400% in early 2025.

👉 Read IDlayr's analysis of AI fraud, mobile possession, and identity trust

Context

AI-enabled fraud changes the identity problem by making the attacker cheaper, faster, and more convincing than the controls built around human verification. Passwords, PINs, SMS OTPs, voice checks, and document review all depend on a person spotting deception or protecting a secret, and that assumption is now under pressure across consumer identity, KYC, and payment journeys.

The core governance issue is no longer only fraud volume. It is that the trust model behind many authentication flows still assumes a human can safely mediate verification, while the article argues that possession-based controls shift the decision to a device or network layer where AI cannot easily copy or intercept the factor.

For identity programmes, that means authentication strategy, fraud controls, and mobile trust now overlap more tightly than many organisations have planned. The article’s starting position is typical of the market shift, but not yet typical of enterprise control design.

Key questions

Q: How should security teams reduce AI-enabled account takeover risk in authentication flows?

A: They should remove reliance on user-entered secrets wherever the attacker can impersonate the conversation or the interface. The most effective move is to shift high-risk steps to possession-based verification tied to a physical device or SIM, then reserve human-mediated checks for lower-risk recovery paths and exception handling.

Q: Why do SMS OTP and voice checks fail more often against AI-driven fraud?

A: Because both depend on a person recognising deception and acting correctly under time pressure. AI can clone voices, generate convincing prompts, and relay one-time codes in real time, which turns a supposedly short-lived secret into an interceptable artefact. That makes the control fragile even when users are careful.

Q: What do organisations get wrong about possession-based authentication?

A: They often treat it as a nicer user experience rather than a structural trust change. The key difference is that possession-based authentication moves verification to a device or network layer, where the attacker cannot simply copy the factor from a prompt, a text message, or a synthetic voice call.

Q: How do teams decide when to use mobile network verification instead of human challenge steps?

A: Use mobile network verification when the journey is high value, easily phished, or likely to attract automated fraud. It is most appropriate where the organisation needs a silent proof of possession and where asking the user to confirm identity would create more exposure than assurance.

Technical breakdown

Why knowledge-based authentication fails against AI fraud

Knowledge-based authentication depends on information a person knows, such as passwords, PINs, OTPs, or security questions. AI changes the economics of attacking those factors by scaling phishing, deepfake voice calls, and real-time relay attacks that can capture and reuse secrets before they expire. Once the secret is exposed to a fake interface or a convincing caller, the underlying control no longer distinguishes the attacker from the user. The result is not just more fraud, but lower-cost fraud that can be repeated at machine speed.

Practical implication: treat knowledge factors as high-friction convenience, not as resilient proof of identity in AI-heavy fraud paths.

Why possession factors are structurally harder to fake

Possession-based authentication shifts trust to something the user physically holds, such as a mobile device or SIM card, and verifies that possession at the device or network layer. The security advantage is structural: AI can replicate text, voice, images, and even documents, but it cannot physically possess the user’s device. That makes possession factors more resistant to phishing and synthetic content attacks, especially when the verification happens silently and server-side. The model matters most when authentication needs to survive repeated attempts from automated adversaries.

Practical implication: prioritise factors that are bound to a specific device or network state rather than factors a user can read, type, or repeat.

How mobile network verification changes the trust chain

Silent network authentication turns the mobile network into part of the identity signal, so the enterprise checks whether the SIM and device relationship still matches the expected user. That removes the human from the verification step and replaces an error-prone interaction with a deterministic yes or no response. In practical terms, the control is useful where SMS OTP and voice checks create too much exposure to interception, relay, or coercion. It also supports a stronger chain of trust for mobile-first identity journeys and for emerging agent-led commerce flows that still need a verified human origin.

Practical implication: use network-level possession checks where the business currently relies on users to prove they are legitimate by entering codes or answering prompts.

Threat narrative

Attacker objective: The attacker wants to bypass identity checks cheaply at scale, then complete account takeover, fraudulent onboarding, or payment abuse without needing the victim’s ongoing participation.

1. Entry begins when an attacker uses AI-generated phishing, voice cloning, or fake documents to reach a user’s authentication flow and capture a knowledge factor or session path.
2. Escalation follows when the attacker relays or reuses the stolen factor in real time, or when the human is tricked into approving a transaction or enrolment that should not have been trusted.
3. Impact occurs when the attacker authenticates as the victim, takes over the account, or completes fraudulent onboarding and payment activity that appears legitimate to the system.

Breaches seen in the wild

• MITRE ATT&CK Enterprise Matrix — MITRE ATT&CK Enterprise — adversary tactics and techniques, threat detection, attack chain mapping, credential access, lateral movement, privilege escalation.
• Code Formatting Tools Credential Leaks — Widely used code formatting tools cause massive credential and secrets leaks in enterprise environments.

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

NHI Mgmt Group analysis

Knowledge factors have become a liability when AI can industrialise deception. Passwords, PINs, OTPs, and security questions were designed for a world where attackers had to work one victim at a time. That assumption collapses when AI can generate credible lures, relay captures in real time, and repeat the attempt at scale. The implication is that identity assurance can no longer depend on the user correctly detecting fraud under pressure.

Mobile possession is now the more defensible trust anchor for consumer identity. A possession factor tied to a specific device or SIM is structurally different from a secret a person can repeat or disclose. That does not make mobile the answer to every identity problem, but it does make device-bound proof more resilient in channels where the attacker controls the interaction and the user has seconds to react. Practitioners should treat that distinction as a governance priority, not a feature preference.

Human-in-the-loop authentication is the weak point AI exploits first. The article is right to frame the user as the control surface, because most fraud journeys still ask a person to read, verify, type, or confirm. That interaction model was already fragile before deepfakes and relay kits; now it is a predictable failure mode. For identity leaders, the lesson is that trust should move away from human judgment and toward cryptographic or network-verified signals wherever possible.

Identity and fraud teams need one shared operating model for AI-era attacks. Authentication, fraud detection, KYC, and mobile trust are no longer separable concerns when synthetic content can cross all four layers in a single attack chain. The organisations that still treat these as separate programmes will keep solving for symptoms in one layer while attackers exploit the seams between them. Practitioners should align governance, telemetry, and response around the full identity journey.

From our research:

• The average estimated time to remediate a leaked secret is 27 days, despite 75% of organisations expressing strong confidence in their secrets management capabilities, according to The State of Secrets in AppSec.
• Organisations maintain an average of 6 distinct secrets manager instances, creating fragmentation that undermines centralised control, according to The State of Secrets in AppSec.
• For broader NHI risk framing, review 52 NHI Breaches Analysis for recurring patterns in exposed credentials and access abuse.

What this signals

Mobile possession becomes the right kind of control when the attacker can already imitate the user. AI fraud reduces the value of anything a user can read, repeat, or verbally confirm. That pushes identity programmes toward device-bound signals, carrier-verified possession, and tighter integration between IAM and fraud operations, especially for recovery and step-up journeys.

Identity leaders should expect authentication to split into two governance paths. One path will keep lightweight convenience checks for low-risk interactions. The other will reserve stronger, possession-based proof for enrolment, recovery, and payment-adjacent flows where the cost of a bypass is materially higher. Organisations that do not separate those paths will keep overloading users with brittle controls.

The pressure on human-mediated authentication also raises the value of the 52 NHI Breaches Analysis for understanding how exposed credentials, reused secrets, and poor lifecycle control keep turning identity signals into attack surface. The lesson for practitioners is that stronger proof only matters when the control path is both durable and governable.

For practitioners

• Replace human-mediated OTP paths Move high-risk authentication journeys away from SMS codes, voice callbacks, and knowledge questions where users can be socially engineered or relayed in real time. Prioritise possession-based checks for recovery, enrolment, and step-up flows that currently depend on user discretion.
• Map fraud journeys to the identity control surface Trace where the user is asked to read, type, approve, or verbally confirm identity and mark each step as an attacker opportunity. Use that map to remove manual judgment from the highest-value flows first, especially where account takeover or payment abuse would have immediate impact.
• Adopt device- and network-bound proof Prefer factors that verify the physical device or SIM at the network layer so the check cannot be copied from a prompt or replayed from a fake site. This is most useful when the business needs silent verification that does not rely on a user spotting a phishing attempt.
• Align IAM and fraud teams on one trust model Create a shared policy for when authentication should escalate from knowledge factors to possession factors and define ownership for exceptions, recovery, and telemetry review. If fraud and IAM disagree on what counts as proof, attackers will keep using the gap.

Key takeaways

• AI has made knowledge-based authentication economically weak because attackers can now clone voices, relay codes, and generate believable documents at scale.
• Possession-based controls are more resilient because the attacker must prove physical control of a specific device or SIM, not just imitate a user-facing signal.
• Identity and fraud programmes should converge on the same trust model, with human-mediated steps removed from the highest-risk authentication journeys.

Key terms

• Possession-Based Authentication: An authentication method that proves the user has a specific physical device, SIM, or security token. In practice, it shifts trust away from information a person can share or forget and toward a factor that is harder for AI-driven fraud to imitate or intercept.
• Silent Network Authentication: A verification method that confirms a mobile identity through a secure interaction with the mobile network, without asking the user to type a code or make a choice. It is especially relevant where user-mediated steps are too easy to phish, relay, or coerce.
• Human-In-The-Loop Fraud: A fraud pattern that depends on a person to confirm, repeat, or approve a request at the wrong moment. AI makes these attacks more effective because it can create convincing prompts, voices, and documents that exploit human judgment rather than bypassing it outright.
• Phishing-Resistant Authentication: An authentication approach that does not rely on the user transmitting a reusable secret to prove identity. The control reduces exposure to fake websites, relay attacks, and social engineering because the proof is verified in a way the attacker cannot easily steal and replay.

What's in the full article

IDlayr's full article covers the operational detail this post intentionally leaves for the source:

• A deeper explanation of silent network authentication and how the verification flow works across the mobile network and device.
• The FAQ discussion of why mainstream passkeys behave differently from true device-bound possession factors.
• Examples of how possession-based authentication can replace SMS OTP in recovery and step-up journeys.
• The article’s framing for agent-led commerce and why verified identity links matter when transactions become partially automated.

👉 The full IDlayr article covers the mobile network verification model, agentic commerce risks, and FAQ detail on passkeys.

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 identity security programme, it is worth exploring.