AI fraud now needs both identity and session authenticity controls

At a glance

What this is: Incode says AI fraud has outgrown identity-only verification because attackers can now defeat both document checks and the session itself.

Why it matters: This matters to identity, fraud and IAM teams because verification controls must now cover both who the user is and whether the interaction is genuine.

By the numbers:

• 40% of legitimate users abandon document-based verification flows before completion, according to Incode.
• GovFaceMatch verifies legitimate users in just 10 seconds, on average, according to Incode.
• Deepsight and GovFaceMatch together deliver a 20% improvement in overall conversion rate, according to Incode.

Context

Identity verification has usually tried to prove a person is real and then stop there, but that model breaks down when attackers can fabricate both identities and the live session around them. The governance problem is no longer just proofing at enrolment, it is verifying that the interaction is authentic throughout the transaction.

This is where identity verification intersects with fraud governance and, increasingly, IAM-adjacent trust controls. If the session can be faked with synthetic video, rooted devices or automated bots, then traditional identity checks become only one control layer, not the whole answer.

Key questions

Q: What breaks when identity verification assumes there is always a human behind the session?

A: The model breaks when the subject is an AI agent or a synthetic identity that cannot be verified through conventional human attributes. In that case, document checks, facial matching, and personal knowledge stop being reliable anchors for trust. Governance has to shift toward ownership, scope, and session-level authorisation.

Q: Why do AI-generated fake IDs and deepfakes create such a sharp fraud risk in digital onboarding?

A: AI-generated documents and biometric spoofing reduce the quality gap that traditional verification once depended on. When attackers can create realistic passports, IDs, faces, and supporting data at scale, basic KYC controls may accept fraudulent identities as legitimate. The risk is highest where onboarding is fast, review is shallow, and verification is not tuned to detect synthetic patterns across documents, devices, and user behaviour.

Q: What are the warning signs that identity verification is too easy to spoof?

A: The main warning signs are heavy reliance on a single human-recognition cue, fast approvals for high-risk requests, and recovery workflows that can be completed during one conversation. If the process can be driven by tone, familiarity, or urgency, it is too easy to spoof.

Q: How should fraud and IAM teams decide whether to prioritise friction or stronger assurance?

A: They should prioritise whichever combination reduces overall risk without shifting the burden onto legitimate users. The decision should be based on conversion, fraud capture and assurance quality together, not on friction alone. If the architecture can verify both the person and the session in one flow, that usually outperforms adding extra manual steps.

Technical breakdown

Why identity checks fail when the session is synthetic

Traditional IDV compares a presented credential or selfie against reference data and then assumes the interaction is genuine. That works only when the attacker cannot tamper with the capture channel. Deepfake injection attacks change the premise by replacing the live camera feed or automating the interaction, so the system sees consistency even when the user, device and video source are not trustworthy. In practical terms, the attack bypasses liveness checks by attacking the session layer rather than the identity artefact. This is why document checks and basic selfie matching no longer provide enough assurance on their own.

Practical implication: Practitioners need controls that validate the capture session itself, not just the identity document or biometric match.

How AI-powered identity fraud exploits stolen PII and fabricated documents

Modern fraud workflows increasingly start with real identity data that has already leaked. Attackers combine stolen personally identifiable information with high-quality synthetic documents and their own face, creating a plausible chain for systems that treat each signal separately. The result is not a broken signature or an obviously fake document, but a coherent set of inputs that passes probabilistic checks. That makes the control failure one of correlation quality: if the system cannot bind the presented face, document and authoritative record tightly enough, synthetic identities can move through onboarding at scale.

Practical implication: Teams should evaluate whether their proofing flow binds biometric, document and authoritative-record evidence tightly enough to resist synthetic identity construction.

What session authenticity adds to identity verification

Session authenticity checks whether the interaction is occurring in real time on a real device with an unmanipulated channel. That means looking for biometric depth signals, behavioural anomalies and device or camera integrity failures that indicate virtual camera injection or bot activity. This is structurally different from a pure identity check because it is focused on the transaction context, not only on the identity claim. In fraud terms, it closes the gap between a valid identity assertion and a trustworthy live session, which is where many AI-enabled attacks now operate.

Practical implication: Security and fraud teams should treat session integrity as a first-class control surface alongside identity proofing.

Threat narrative

Attacker objective: The attacker wants to pass verification controls while looking like a real user, enabling fraud, account takeover or synthetic enrolment without triggering friction.

1. Entry begins with stolen personally identifiable information and synthetic documents that make the target look legitimate to identity systems.
2. Credential and session abuse follow when attackers use AI-generated media, virtual cameras or rooted devices to present a convincing live interaction.
3. Impact occurs when the platform accepts both the identity claim and the manipulated session, allowing fraudulent enrolment or account access at scale.

Breaches seen in the wild

• Arup deepfake fraud 2024: Deepfakes of Arup's CFO and colleagues on a video call led a Hong Kong employee to transfer HK$200 million (about US$25.6m) to fraudsters.

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

Session authenticity is now part of identity governance, not an optional fraud add-on. Once attackers can fake the live interaction, proofing controls that stop at the document or selfie layer leave a governance gap. That gap matters because the system may be verifying evidence, not trust. Practitioners should treat the session as an identity control boundary, not just a transport detail.

AI has collapsed the old conversion-versus-fraud tradeoff into a control design problem. Organisations that try to solve deepfake risk by adding more friction are usually shifting risk onto legitimate users rather than removing it. The better question is whether the verification architecture can raise assurance without forcing extra steps. The useful concept here is session-aware identity assurance, which binds proofing to the real-time interaction and reduces the room for synthetic inputs.

Identity verification without authoritative source binding is increasingly easy to imitate. The article's most important implication is that local comparisons between selfie and document are too easy to stage once attackers control the media. Matching against a trusted issuing source changes the control model because it reduces dependence on user-submitted artefacts. Practitioners should re-evaluate whether their current proofing model is anchored to an authoritative record or still trusting the applicant's own presentation.

Deepfake detection and identity proofing are converging into a single governance decision. Fraud teams, IAM leads and security architects can no longer treat liveness, device integrity and identity proofing as separate programmes with separate thresholds. The operational question is which evidence sources are allowed to establish trust and how much of the session must be verified in real time. Practitioners should align fraud controls with identity assurance policy rather than manage them as disconnected tools.

From our research library:

• Gartner predicts that by 2026, 30% of enterprises will consider identity verification solutions unreliable in isolation because of AI-driven attacks.

What this signals

Session-aware identity assurance: the next control boundary is no longer the document image or selfie alone, but the integrity of the live interaction that carries them. Fraud programmes that do not account for synthetic video, virtual camera injection and bot mediation will keep overestimating the value of visual checks.

For identity and fraud teams, the practical shift is to design for authoritative-source binding and real-time session validation at the same time. That means the onboarding path has to prove both the identity claim and the trustworthiness of the channel before approval is granted.

For practitioners

• Map proofing controls to both identity and session assurance Review onboarding and step-up journeys to separate identity evidence from session integrity evidence, then define which controls prove each one.
• Reduce reliance on user-submitted artefacts Limit the weight given to uploaded document images when a stronger authoritative source or biometric binding is available.
• Add session-integrity checks to fraud workflows Use device integrity, camera integrity and behavioural signals to detect virtual camera injection, bot activity and manipulated video feeds.
• Revisit conversion metrics alongside fraud loss Measure abandonment, approval rate and fraud capture together so friction decisions reflect total control performance rather than a single KPI.

Key takeaways

• AI-enabled fraud now succeeds by attacking both identity evidence and the live session, so a single verification layer is no longer enough.
• The article's key evidence is that document-based flows still lose legitimate users, while deepfake-driven impersonation attempts are rising sharply.
• The strongest control response is to bind identity proofing to authoritative records and add session-integrity checks that detect manipulated interactions.

Key terms

• Session Authenticity: The assurance that a live verification session is genuinely created by a real user on a real device at the time of the check. It matters because attackers can manipulate video, cameras, or automation layers without changing the identity data itself. Strong session authenticity controls look for provenance, integrity, and replay resistance.
• Authoritative Source Binding: Authoritative source binding is the practice of verifying identity evidence against a trusted issuer or system of record instead of relying mainly on user-submitted artefacts. It reduces dependence on documents that can be fabricated, altered or paired with stolen personal data.
• Deepfake Injection Attack: A fraud technique that replaces or alters the live input stream during a verification process with synthetic media. The target is not always the identity record. The attacker may instead attack the session layer by injecting fake video, camera output, or replayed content that appears authentic enough to pass checks.
• Identity Assurance: The confidence an organisation has that a person or system is truly who it claims to be before access or action is granted. In modern IAM, assurance depends on evidence quality, channel trust, and the strength of verification around high-risk decisions.

Deepen your knowledge

The NHI Foundation Level course, the industry's only accredited NHI security programme, covers NHI governance, identity lifecycle and machine identity security. It gives practitioners a practical way to connect identity controls to the wider security programme they run.