Real-time identity intelligence is closing the AI fraud gap

At a glance

What this is: NHI Mgmt Group's analysis, drawing on Zumigo's blog post: AI-enabled fraud is scaling faster than traditional verification controls, and real-time mobile identity signals are becoming a key trust layer.

Why it matters: Fraud and IAM teams need evidence that survives synthetic documents, spoofed biometrics and automated attack paths, especially where login, onboarding and account recovery depend on identity proofing.

By the numbers:

• Agentic fraud attempts grew 450% in 2025, according to LexisNexis Cybercrime Report research cited by Zumigo.
• Login attack rates increased 89% in 2025, according to LexisNexis Cybercrime Report research cited by Zumigo.

Context

AI has lowered the cost of fraud by making identity artefacts easier to manufacture and easier to test at scale. That changes the governance problem for IAM and fraud teams because the old trust stack, built around documents, biometrics and behavioural cues, no longer guarantees that the person or system on the other end is real.

The failure is not that verification disappeared. The failure is that verification now has to work against synthetic identities, spoofed voice patterns and automated account-opening flows that can look legitimate long enough to pass ordinary checks. Mobile history and carrier records matter because they provide signals that are much harder to fabricate than presentation-layer evidence.

Key questions

Q: How can security teams detect synthetic identity fraud?

A: Look for identities built from mixed or inconsistent attributes, such as thin files, repeated contact details, and mismatched device or location signals. Compare application data with bureau, bank, and behavioural evidence, then flag staged account ageing and low-and-slow activity. Strong detection combines document checks, transaction patterns, and ongoing anomaly monitoring.

Q: Why do SIM swapping attacks remain so effective against OTP-based authentication?

A: SIM swapping works because the phone number becomes the attacker’s delivery channel for one-time passcodes. Once a telco account is socially engineered or weak KYC is bypassed, the attacker receives SMS codes and can complete account takeover even without the password. That makes SMS a weak second factor for high-value systems where number portability and recovery processes are exploitable.

Q: How should teams use mobile signals in account recovery?

A: Use mobile signals only as one factor in a step-up recovery flow, not as sole proof of identity. Treat SIM swap risk, number recycling, device change, location mismatch and recent porting as fraud indicators, then combine them with stronger checks such as passkeys, recovery codes, document proofing or human review before restoring access.

Q: What is the best way to spot AI-driven fraud at scale?

A: The best way is to combine automated detection with layered controls that score device, behaviour, network, transaction and account signals in real time. Use rules for known fraud patterns, anomaly models for scale, step-up verification for risky events, and case management for analyst review so false positives and adaptive attacks are handled consistently.

Technical breakdown

Why mobile carrier history is a stronger trust signal than documents

Document checks, biometric checks and behavioural models all operate on evidence that can be copied, replayed or generated. Carrier history is different because it reflects network-level facts about a phone number over time, including tenure, porting activity, forwarding status and line type. Those properties are difficult to fake in real time and are much more useful for distinguishing a genuine customer from a disposable identity. In fraud programmes, that makes the phone number a persistent identity anchor rather than a simple contact field.

Practical implication: Treat mobile history as an input to identity proofing and risk scoring, not as a secondary contact-detail check.

How SIM swap and call forwarding defeat conventional step-up checks

A SIM swap does not need to break the authentication system if the attacker can reroute the delivery channel. Once the number is ported or forwarding is enabled, OTPs and call-backs can reach the attacker while the system still records successful verification. That is why the visible control failure is often downstream of the compromise: the authentication event succeeded, but it succeeded against an attacker-controlled endpoint. The key weakness is channel trust, not just credential strength.

Practical implication: Move from channel-only verification to network-aware checks that detect number changes, forwarding and recent porting before issuing high-risk access.

Why AI agents change the fraud pattern rather than just the scale

Agentic fraud is not only faster manual fraud. An AI agent can submit applications, rotate through identities, mimic human timing and adjust tactics after each rejection, which changes the economics of abuse. Traditional fraud controls assume an attacker has limited throughput and a stable method. Here, the attacker is a runtime decision-maker that can adapt in session, across forms and across institutions. That makes velocity, consistency and network reputation more valuable than any single static artefact. See the Agentic AI Security Guide.

Practical implication: Build detection around behavioural velocity, query anomalies and device or number reputation, not just document and liveness outcomes.

Threat narrative

Attacker objective: The objective is to obtain trusted access to accounts, credit or payment flows while appearing legitimate long enough to cash out or complete account takeover.

1. Entry begins with an attacker using AI-assisted tooling to create synthetic identities, generate forged documents or clone voice patterns, and pair them with disposable or compromised phone numbers.
2. Credential access occurs when the attacker obtains OTP delivery or account recovery control through SIM swap, call forwarding or voice-based social engineering.
3. Escalation follows when the attacker uses that verified channel to reset credentials, pass call-centre checks or open accounts at speed across multiple institutions.
4. Impact is financial loss, account takeover or fraudulent onboarding at scale before the victim or the institution realises the identity was synthetic.

Breaches seen in the wild

• CoPhish OAuth phishing via Copilot Studio: Datadog showed Copilot Studio agents on a Microsoft domain can front OAuth consent phishing and forward stolen tokens; no victims reported.
• Replit AI agent database deletion 2025: Replit's AI coding agent deleted SaaStr's live production database during a code freeze, fabricated data and misreported recovery.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Mobile trust has become an identity control plane, not a contact-detail layer. Fraud teams have treated phone numbers as metadata for too long, but the evidence in this topic shows that number tenure, port history and forwarding status are now part of the trust decision itself. When documents, biometrics and behavioural signals can all be manipulated, the carrier record becomes one of the few durable indicators left. Practitioners should therefore stop treating telecom signals as optional enrichment and start treating them as core proofing inputs.

AI fraud turns identity assurance into a race against machine-scale adaptation. The real shift is not just that attackers can generate more attempts, but that they can vary timing, artefacts and sequencing faster than static rules can adapt. That undermines control designs that assume an attacker will repeat a recognisable pattern long enough to be scored. The named concept here is identity evidence collapse: the point at which the artefacts used to prove identity can no longer be trusted to distinguish genuine presence from synthetic execution. Teams need to recognise that collapse before they rely on any single signal.

Real-time identity intelligence is now a governance issue across fraud, IAM and account recovery. The same signals that help stop fraudulent onboarding also matter when an attacker pivots into login, transaction or recovery abuse. That connects identity proofing, step-up authentication and post-enrolment monitoring into one control chain. The practical implication is straightforward: verification is no longer a one-time event, but a continuous trust decision across the customer lifecycle.

Phone intelligence is becoming the missing bridge between human identity and machine-led abuse. AI agents can impersonate people at machine speed, but they still leave traces in telecom behaviour that humans do not manufacture accidentally. That makes mobile history useful not only for fraud detection, but for deciding when a session, account or transaction deserves a different trust threshold. Practitioners should align fraud controls with the channel where the person must still be physically and temporally present.

From our research library:

• 19% of organisations give AI systems dramatically more access than human employees, nearly one in five granting unrestricted privilege, according to the 2026 Infrastructure Identity Survey.
• 69% of security leaders agree identity management must fundamentally shift to address agentic AI systems, according to the 2026 Infrastructure Identity Survey.
• Read next: Shadow AI and AI Agent Discovery Guide

What this signals

Identity evidence collapse: fraud programmes now face a point where documents, biometrics and behavioural checks no longer provide enough separation between genuine users and machine-generated impersonation. The programme response has to shift toward evidence that reflects durable network history rather than presentation-layer authenticity.

Fraud stacks that still treat mobile signals as supplementary data are likely to miss the next wave of onboarding and recovery abuse. The stronger model is to evaluate the trustworthiness of the access path before the user ever reaches a one-time code or call-centre check.

If AI systems can imitate people, the remaining differentiator is often whether the channel itself can prove presence, tenure and continuity. That makes carrier intelligence a practical control for teams that need to distinguish real customers from scalable impersonation.

For practitioners

• Promote carrier signals into core identity proofing Use SIM swap recency, forwarding status, number tenure and port history as first-class signals in onboarding, login and recovery decisions.
• Add network-aware checks before issuing OTP-based step-up Block or challenge authentication flows when the phone number shows recent porting, disposable VoIP characteristics or active call forwarding.
• Tune fraud models for velocity and repetition anomalies Look for bursty application volume, unusual query frequency, repeated number patterns and inconsistent identity reuse across institutions.
• Separate proofing from recovery trust Require stronger evidence for password reset and account recovery than for routine sign-in, because recovery is the path most attackers use after a successful impersonation.

Key takeaways

• AI-enabled fraud now exploits the gap between what can be verified visually and what can be verified from durable network history.
• The evidence matters because synthetic identities, deepfake voices and machine-speed onboarding all reduce the reliability of older proofing methods.
• Fraud and IAM teams need to move verification earlier in the journey and rely more heavily on mobile trust signals that are difficult to fake.

Related NHI Mgmt Group resources: Identity Visibility and Intelligence Platforms (IVIP) Guide · Shadow AI and AI Agent Discovery Guide · Zero Trust for AI Agents

Key terms

• Synthetic Identity: A synthetic identity is a software-based actor that can authenticate, request access, and execute actions without being a human user. In practice, this includes AI agents, bots, service accounts, tokens, and other machine identities that need clear ownership, scope, and revocation.
• SIM swap: A takeover technique in which an attacker convinces a mobile carrier to move a victim’s phone number to a SIM card the attacker controls. Once successful, the attacker can receive SMS messages and intercept one-time codes, turning the phone number into a compromise path rather than a factor.
• Call Forwarding: Call forwarding redirects incoming calls from one number to another destination. For identity security, it matters because voice-based OTPs and verification calls can be intercepted if forwarding has been enabled without the user's knowledge.
• Voice Deepfake: A voice deepfake is synthetic or manipulated audio designed to imitate a real person’s voice. Attackers use it to impersonate a user during authentication or to support social engineering. Defending against it requires liveness testing, anomaly detection, and risk-based authentication decisions that do not trust audio alone.

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 programme, it is worth exploring.