EU AML rules tighten identity verification for digital onboarding

At a glance

What this is: NHI Mgmt Group's analysis, drawing on iProov's blog post: The EU's new AML rules will reshape digital onboarding by making eID the default and exposing the limits of legacy identity verification.

Why it matters: IAM, IGA, and fraud teams need to understand that onboarding controls are becoming a regulated assurance problem, not just a user experience choice.

By the numbers:

• The new EU anti-money laundering regulations come into force on 10th July 2027.
• A single bad actor opened 46 bank accounts with ABM AMRO using stolen identity information and deepfakes.

Context

The governance gap here is not whether organisations can collect identity data, but whether they can verify identity to a standard that still holds up against deepfakes and injection attacks. The article frames the issue as a shift from flexible onboarding practice to regulated assurance, with eID becoming the default control for EU digital onboarding.

For IAM and fraud teams, the important change is that onboarding is moving from a documentation exercise to a risk-managed identity proofing model. That affects customer due diligence, account opening, and the evidence an institution needs to justify any fallback to physical documents or video checks.

Key questions

Q: What breaks when remote onboarding still relies on document and selfie checks?

A: The control breaks when the verification path assumes visible evidence is enough to prove identity. Deepfakes, forged documents, and injection attacks can all satisfy a weak front-end check while a fraudulent customer still enters the KYC perimeter. The failure is not onboarding itself, but the trust model behind it.

Q: Why do stronger identity proofing controls reduce AML risk?

A: Because AML controls only work if the identity behind the account is real enough to be trusted. When criminals can open accounts with stolen or synthetic identities, the institution inherits money laundering, sanctions, and fraud exposure from the start. Stronger proofing raises the cost and complexity of creating believable false identities.

Q: How can security teams tell if onboarding controls are not providing real assurance?

A: A weak onboarding programme shows up when fraud is only discovered after interview completion, device issuance, or access grant. If manual review, background checks, and document scans still allow impostors through, the control is validating paperwork rather than identity.

Q: What should organisations do when eID is required but not yet certified in time?

A: They should treat certification timing as a deployment risk and not a procurement afterthought. If an onboarding path cannot be certified before the deadline, it should not be counted on for the regulated flow. The safer choice is to redesign the onboarding route early enough to avoid a last-minute compliance gap.

Technical breakdown

Why eID becomes the default onboarding control

The draft RTS changes onboarding from a local policy choice into a regulated identity assurance workflow. eIDAS provides the assurance framework, and the AML package pushes institutions to use electronic identity verification as the primary route rather than treating it as one option among many. In practice, that means the institution must be able to map onboarding methods to assurance levels and explain why a fallback was used. The control problem is not just identity proofing, but proving that proofing met the required level for the transaction and customer risk profile.

Practical implication: align onboarding flows to eID assurance levels now, before supervisors expect documented justification for every fallback.

How deepfakes and injection attacks bypass legacy verification

Legacy remote identity checks often assume that comparing a face, a document, and a live session is enough to establish trust. The article shows why that assumption is weakening: presentation attacks can fool visible checks, while injection attacks bypass the camera path entirely by feeding synthetic media into the stream. That makes basic presentation attack detection insufficient when attackers can manufacture a plausible identity session end to end. The technical issue is not just spoofing, but the ability to inject false evidence into the verification pipeline before the control can distinguish real capture from fabricated input.

Practical implication: treat camera-path integrity and injection resistance as separate requirements, not as an extension of document or selfie checks.

Why assurance level and certification timing now matter

The article makes clear that compliance risk is tied to certification lead time as much as to control design. If a vendor or internal stack is not yet certified to the required eIDAS assurance level, an organisation may not be able to deploy it safely before the regulatory deadline. The operational problem is that identity assurance is now a procurement and lifecycle issue, not just a technical one. This shifts onboarding controls into the same planning category as other regulated access systems where evidence, assurance, and deployment timing must line up.

Practical implication: review certification status and implementation lead times together, because a late-control purchase can become a compliance failure.

Threat narrative

Attacker objective: The attacker wants to create fraudulent accounts that appear legitimate enough to pass onboarding controls and support financial crime activity.

1. Entry occurs when attackers use stolen identity data, deepfakes, or forged documents to begin the remote onboarding attempt.
2. Credential or proof access succeeds when the attacker bypasses the document check or injects synthetic media into the verification stream.
3. Escalation follows when the fraudulent identity is accepted inside the KYC perimeter and can be used to open accounts at scale.
4. Impact is realised through money laundering, sanctions evasion, or other abuse of accounts created with weak onboarding assurance.

Breaches seen in the wild

• Spain's first AI agent data breach 2026: Spain's AEPD logged its first breach notification attributed to an attacker's AI agent, which altered personal data and accessed invoices.

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

The regulatory centre of gravity is shifting from identity collection to identity assurance. EU AML onboarding is no longer mainly about gathering enough customer data to satisfy a form. It is about proving that the verification method itself can withstand modern fraud tactics and still support risk-based due diligence. For practitioners, that means onboarding controls must be judged by assurance quality, not by convenience or legacy habit.

Document and selfie checks are becoming governance liabilities when injection attacks are in scope. The article’s distinction between presentation attacks and injection attacks matters because many organisations still treat them as the same problem. They are not. Injection attacks undermine the capture layer itself, which means a process can look compliant while still accepting fabricated evidence. The practical conclusion is that onboarding governance must account for capture integrity, not only identity attributes.

Identity proofing is now a lifecycle and certification problem, not a point-in-time project. The article highlights certification cycles, regulatory timing, and assurance levels in the same discussion because they now move together. That combination means onboarding capability cannot be procured as a static control and left untouched. For security and compliance leaders, the governance question is whether the identity stack can remain certifiable while the threat model keeps changing.

Identity assurance debt is the new onboarding blind spot. This is the gap created when organisations continue using controls that were adequate for earlier fraud patterns but are no longer sufficient against AI-generated content and injection attacks. The debt accumulates when institutions defer certification, rely on fallback processes too often, or treat eID adoption as a future-state project. Practitioners should see onboarding as an assurance posture to be managed continuously, not a box to be checked once.

Risk-based onboarding is becoming the only defensible operating model. The article’s emphasis on matching assurance level to transaction, identity, and context shows where the market is heading. Institutions will need to make and document narrower, more explicit decisions about when high assurance is required and when lower assurance can be justified. That is a governance shift, not just a tooling change.

From our research library:

• Organisations maintain an average of 6 distinct secrets manager instances, creating fragmentation that undermines centralised control, according to the State of Secrets in AppSec.
• Gartner predicts that by 2026, 30% of enterprises will consider identity verification solutions unreliable in isolation because of AI-driven attacks.
• Read next: Ultimate Guide to NHIs — Regulatory and Audit Perspectives

What this signals

Onboarding governance is moving toward a stricter assurance model in which the proof method matters as much as the identity evidence itself. Institutions that cannot distinguish between presentation resistance and injection resistance will keep passing users through controls that look modern but fail under AI-assisted fraud.

Identity assurance debt: this is the accumulation of onboarding controls that once looked adequate but no longer match the threat model. Organisations maintain an average of 6 distinct secrets manager instances, creating fragmentation that undermines centralised control, according to the State of Secrets in AppSec, and the same pattern of fragmentation can appear in identity proofing stacks when assurance decisions are spread across disconnected tools.

For financial institutions and crypto businesses, the practical signal is that eID adoption, certification readiness, and anti-spoofing evidence now belong in the same governance conversation. The institutions that align those three elements early will have a clearer path to compliant onboarding when the 2027 deadline arrives.

For practitioners

• Map onboarding paths to eID assurance levels Inventory every digital onboarding route and record which ones can support eID at Substantial or High assurance, then identify where fallback methods are still in use without justification.
• Separate presentation and injection attack controls Test whether your verification stack can detect synthetic media entering the data stream, not just forged documents or replayed faces visible to the camera.
• Review certification lead times now Compare current vendor certification status with the July 2027 deadline and remove any deployment path that cannot be proven ready in time.
• Document fallback justifications for supervisors Create a repeatable decision record for every case where physical documents or video verification are used instead of eID, including the risk reason and approver.

Key takeaways

• The article shows that EU AML onboarding is shifting from document-centric verification to regulated identity assurance anchored to eIDAS.
• It also shows that deepfakes and injection attacks are undermining legacy remote verification controls, especially where basic selfie checks still carry too much trust.
• The practical implication is to treat certification status, fallback governance, and capture-path integrity as one onboarding programme, not three separate tasks.

Key terms

• Identity proofing: The process of verifying that a person is who they claim to be before granting or restoring access. In higher-risk recovery paths, proofing can include stronger evidence checks such as government ID validation or liveness-based facial verification so the assurance level matches the sensitivity of the request.
• Electronic Identification: A regulated way of proving a person or organisation’s identity for digital transactions. In eIDAS 2 contexts, it is not just a login mechanism. It is an assurance process that must support interoperability, auditability, and appropriate privacy handling across jurisdictions and service providers.
• Injection attack: An attack that inserts synthetic or manipulated data directly into the verification flow rather than fooling the sensor itself. For identity programmes, this is a control-path problem, because the attacker may bypass the visible presentation layer and exploit the software decision point.
• Assurance Level: An assurance level is the degree of confidence an organisation has that an identity proofing or authentication outcome is accurate. Higher assurance usually means stronger checks, more evidence, and more governance overhead. The key is matching assurance to the transaction risk, not applying one standard everywhere.

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 responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.