As of July 10, 2027, Photo-Ident will no longer be a viable standard process. The EU Anti-Money Laundering Regulation (AMLR) and its regulatory technical standards establish eID and qualified trust services as the default for remote identification. The underlying standard, ETSI TS 119 461, explicitly disallows an ID card photo (VAL-8.3.3-04X, NOTE 4).
Those who currently onboard users using photos and selfies must transition their standard processes: to automated Video-Ident, ePassport, or eID, and from web browsers to a hardened app.
Three key dates set the timeline:
- July 2027: The AMLR becomes directly applicable across the entire EU.
- August 2027: ETSI TS 119 461 V2.1.1 becomes the benchmark standard for identity verification underlying every qualified electronic signature (QES).
- December 2027: Banks must accept the EUDI Wallet whenever requested by customers.
Table of Contents
Why this applies to every bank
- AMLR, Art. 22(6): For identity verification, the regulation specifies only two digital methods alongside the physical presentation of an identity document: eID means at assurance levels “substantial” or “high”, and qualified trust services such as QES.
- AMLA Regulatory Technical Standards: The final draft dated September 30, 2026 establishes these two methods as the standard rule for remote channels. Alternative remote procedures remain permissible only for customers who lack access to them, and the bank must substantiate such exceptions to the supervisory authority. The draft has been submitted to the European Commission for adoption.
- eIDAS with Commission Implementing Regulation (EU) 2025/1566: Prior to issuing a qualified certificate for a QES, the identity must be verified in accordance with ETSI TS 119 461 V2.1.1, Annex C.3. This represents the “Extended” level.
Why Photo-Ident falls short
ETSI TS 119 461 V2.1.1 does not permit a static photo of an identity document, neither at the Baseline nor at the Extended level. The standard explicitly states:
VAL-8.3.3-04X, NOTE 4: According to the current state of the art, a mere still image of the ID card is not considered sufficient for either of the two levels (“only a still photo of the identity document is not considered sufficient”).
- Real-time (VAL-8.3.3-02X): The identity document must be captured live in front of the camera during identification. Pre-recorded photos or videos are not accepted.
- Video instead of photo (VAL-8.3.3-04X): A video sequence of each relevant page of the identity document is required to make the security features visible.
- Optically variable features (VAL-8.3.3-07X and 07A): At least three security features must be verified, two of which must be optically variable. Holograms and tilt effects only reveal their authenticity through movement.
- Protection against injection and deepfakes (VAL-8.3.3-05A and 05B): The process must repel injected, recorded, or AI-generated videos and detect manipulated documents.
Which AMLR-compliant identification methods are available?
| Method | How it works | Why it is viable |
|---|---|---|
| Automated Video-Ident | The customer records their ID document in motion and captures a selfie video. No video call, no appointment. | AI and trained verifiers validate the document jointly. This corresponds to the ETSI hybrid model (9.2.3.3) and achieves Extended level. |
| ePassport | The customer scans the chip in their passport or ID card via NFC and records a selfie video. | The issuer's digital signature confirms authenticity. ETSI permits this fully automatically (9.2.3.4), likewise at the Extended level. |
| eID | The customer uses the online ID function of their national identity card with a PIN. | eID is a government-issued identification means at assurance level "high", providing a direct path pursuant to Art. 22(6) AMLR. |
| Nect Wallet | Users who already have a Nect Wallet authenticate themselves with a simple selfie video to create a QES. Over 20 million users can already do this today. | The bank identifies the user via QES, one of the two digital methods under Art. 22(6) AMLR. In Germany, QES alone will suffice once reference transfers are eliminated. The Nect Wallet function is already an eIDAS-compliant identification method today. |
Why identification belongs in a hardened app

The attack: replacing the camera
In an injection attack, the fraudster no longer presents physical items to the camera. Instead, they feed a stolen ID photo, pre-recorded video, or deepfake directly into the data stream.
ENISA also warns against this, as such injection attacks no longer require specialized expertise:
- Freely accessible tools: Virtual cameras, such as those included with the streaming software OBS, are not categorized by ENISA as malware, but as standard utilities. The device treats them as legitimate hardware.
- Rapid growth: ENISA identifies injection combined with deepfakes as the prevailing threat today. Gartner observed an increase of 200 percent for 2023. Industry reports identified an increase of 2,665 percent in virtual camera usage in 2024.
Why browsers cannot reliably detect this
- The website only sees self-reported data.
The web standard describes a camera via an identifier, device type, and a textual label. It does not disclose whether a physical sensor or software lies behind it. - The evaluation runs on the attacker’s device.
Both camera and computer are under the attacker’s control. At the highest level, testing laboratories therefore even attempt to modify camera drivers, the browser, or the operating system. - The browser lacks defensive tools.
ENISA highlights: A web process consists solely of the webpage and the browser. A dedicated app enables granular security controls, such as code obfuscation and runtime application self-protection.
Why deepfake detection alone is insufficient
Detection is an arms race. In a benchmark evaluating 2024 deepfakes, the detection rate of freely available video detectors dropped by half.
ENISA considers injection defense to still be in its nascent stages: The techniques are less mature than liveness detection, and even less developed for identity documents than for human faces. Validated countermeasures are expected to require more than five years of continued research.
For this reason, ETSI mandates two separate requirements: detecting deepfakes (BIN-8.4.2-03X) as well as preventing and detecting the injection itself (BIN-8.4.2-04A). The editor of the CEN/TS 18099 testing standard differentiates in the exact same manner: protection against the injection method—such as camera tamper protection, runtime protection, and emulator detection—alongside detection of injected content.
Relying solely on detection requires overcoming every new deepfake model. Preventing the injection cuts off its entry path into the system altogether.
What a hardened app delivers
A hardened app targets the injection at its source: It validates the device and runtime environment, safeguards the camera against tampering, and secures data transmission through to the verification backend. Furthermore, Android and iOS allow an app to cryptographically attest that it is authentic and unmodified.
Even this is not invulnerable. This is precisely why accredited laboratories conduct evaluations using rooted devices and hooking attacks to verify that the defenses hold.
At Nect, this security architecture has proven effective for years: During commissioned penetration tests, no external security firm has successfully bypassed the app hardening. To enable testing of the deepfake detection models, Nect had to deliberately disable the hardening for the testers every time. This demonstrates why the arms race against deepfakes demands a hardened app rather than merely a detection model.
ETSI itself highlights the mobile app: Across three requirements addressing injected material, it notes that compliance is only achievable through software specifically validated for identity verification. Its sole stated example is the mobile app. The standard does not list browsers or any alternative mechanism to prevent injection.

More sensors, stronger defense
- Motion vs. image:
The data registered by the accelerometer must correlate with the visual movement in the video. Researchers leverage this cross-check to detect duplicated and re-recorded screen captures. - Vibration as a watermark:
The system randomly triggers the timing and intensity of a vibration. This leaves physical artifacts in the visual feed, the audio track, and the accelerometer data that are absent from pre-recorded videos. - Audio as a watermark:
The speaker emits a distinct acoustic signal during both the ID and selfie video recordings, including frequencies beyond human hearing. Additionally, the system verifies whether both videos originate from the same acoustic environment.
Nect has obtained European patents for both watermarking techniques: EP 3 575 994 B1 for real-time verification via vibration (granted in 2021) and EP 3 575 993 B1 for the cross-validation of ID and selfie videos (granted in 2022). Both patent specifications also outline the cross-referencing of motion and image data.
An attacker attempting video injection would have to forge image, sound, and motion data synchronously into a coherent sequence they cannot anticipate. This makes the real-time presence required by ETSI (VAL-8.3.3-02X) technically verifiable.
Often, the motion sensor alone provides the most decisive signal. A smartphone held in hand is never completely motionless. If it is entirely static, it is likely mounted on a tripod or resting on a table while a virtual camera injects external material.
In a browser, this signal offers negligible value: A laptop naturally rests motionless on a desk, so stability proves nothing. Furthermore, technical limitations exist: The vibration API is unavailable in Safari and all WebKit browsers, effectively excluding virtually every iPhone. The web standard does not provide motion data as raw telemetry, but rounds it to 0.1 m/s².
In addition, Nect utilizes other smartphone sensors and telemetry to prevent attacks. Which specific parameters are evaluated is deliberately undisclosed to prevent attackers from gaining insights. In a browser environment, this proven defense architecture would not be feasible.
Nect ePass
Caution regarding "ETSI-certified": V1 is not V2
A label indicating “certified according to ETSI TS 119 461” carries little weight without specifying the version and level. Until February 2025, only the V1.1.1 release from 2021 existed, and certificates typically maintain a validity period of two years. Even after V2.1.1 was released, certifications were still being issued under V1.1.1 in 2025, running through 2027.
V2.1.1 imposes significantly stricter requirements on critical aspects:
| Aspect | V1.1.1 (2021) | V2.1.1 (2025) |
|---|---|---|
| Levels | Baseline only | Baseline and Extended. Extended is required for QES. |
| Purely manual verification | Permissible | Only sufficient for Baseline; excluded for QES |
| Security features of the document | Verification "as appropriate", no minimum requirement | At least three, of which two must be optically variable |
| AI-generated or manipulated documents | Not explicitly regulated | Detection is mandatory |
| Injection attacks | Secure transmission only | Defense and detection are mandatory, tested by an accredited laboratory under CEN/TS 18099 |
| Deepfakes in facial video | Detection recommended | Detection is mandatory |
| Liveness detection (PAD) | Evaluation according to ISO/IEC 19989-3 recommended | Evaluation by an accredited laboratory is mandatory, every two years |
A certificate issued under V2.1.1 is not an unconditional guarantee either. Certification is never granted to a provider as a whole, but rather to an individual procedure within a specific use case. The critical factor lies in the scope: If the certificate covers only one product, one use case, or one channel, everything else remains uncovered—even if provided by the same vendor.
Therefore, ensure you obtain verification of five key elements today:
- Version: V2.1.1, not V1.1.1.
- Level: Extended, not Baseline.
- Method and Use Case: The precise procedure you are procuring.
- Channel: App or browser, iOS and Android.
- Laboratory reports: The injection and presentation attack tests mandated by the standard for the first time by the end of 2026.

With Nect: AMLR Ident and EUDI Wallet from a Single API
Nect provides automated Video-Ident, ePass, and eID through a unified interface. All three procedures have been certified under ETSI TS 119 461 V2.1.1 at the Extended level since July 2026. Thus, Nect Ident already delivers all user-friendly and cost-effective identification methods envisioned by the AMLR for the remote channel.
With Nect, you do not need to parse which procedure is covered: Automated Video-Ident, ePass, and eID are all three included in the certificate.
Nect supplies the hardened app: All procedures execute within the Nect Wallet app. Through it, over 20 million users can already open their bank accounts via a simple selfie video, fully compliant with eIDAS. This Wallet-Ident (sometimes termed Re-Ident) has been eIDAS-compliant at Nect for three years, serving as the precursor to EUDI.
EUDI arrives starting in 2027 and will be accessible to all clients integrated with Nect Ident. This enables you to satisfy the mandatory acceptance requirement applicable to banks starting December 24, 2027. For customers without a wallet, Nect automatically routes them to the appropriate alternative identification method.
Legal Framework
- AMLR: Regulation (EU) 2024/1624, Art. 22(6). Applicable from July 10, 2027.
- Regulatory standards: AMLA, Final Report on draft RTS under Art. 28(1) AMLR dated September 30, 2026: Articles 6 and 7, Recital 11. Pending adoption by the European Commission.
- eIDAS: Regulation (EU) No 910/2014 as amended by Regulation (EU) 2024/1183. Art. 24 governs identity verification for qualified certificates. Art. 5f(2) mandates acceptance of the EUDI Wallet upon the user’s request, 36 months after the entry into force of the wallet implementing acts on December 24, 2024. Micro and small enterprises are exempt (Overview by Heuking).
- Reference standard: Commission Implementing Regulation (EU) 2025/1566. Applicable from August 19, 2027.
- ETSI TS 119 461 V2.1.1 (2025-02): Full text. Clauses 8.3.3 and 8.4.2, Use Cases 9.2.3.3 and 9.2.3.4, Annex C.3.
- Nect Ident certification: Announcement of July 22, 2026.
