On 13 October 2026, Switzerland's Federal Administrative Court partly upheld PostFinance's appeal against an order from the Federal Data Protection and Information Commissioner (FDPIC). The court nonetheless agreed with the regulator on the core question. A bank that creates customer voiceprints needs the data subject's express consent. According to the FDPIC's summary of the judgments, a customer who is told about voiceprint creation and simply keeps talking has not given it. The first and most important thing the case settles is which consent standard applies to biometric data.
What the court decided
The court issued two judgments on 13 October 2026. The first, A-4467/2025, concerns the FDPIC's decision of 16 May 2025. That decision told PostFinance AG to obtain the express consent of data subjects when creating voiceprints. The FDPIC announced it on 7 August 2025, and PostFinance appealed. Digital Policy Alert's record of the ruling describes the outcome as a partial win for the bank, because the court extended the deadline for existing voiceprints. Per the FDPIC, PostFinance now has until mid-February 2027 to obtain explicit consent for voiceprints already created. Where no consent is obtained, the data must be deleted. Netzwoche reports that PostFinance has used voiceprints since September 2018.
The second judgment, A-6813/2025, concerns publication. The court confirmed that by publishing its decisions the FDPIC is fulfilling its statutory duty to inform the public. The FDPIC says it will publish the 16 May 2025 decision once that judgment becomes final.
Both judgments apply the revised Federal Act on Data Protection (nFADP). I did not review the judgment texts. This account rests on the FDPIC's published summary, Digital Policy Alert and press coverage. Netzwoche also reports that the court did not assess whether PostFinance's current procedure complies, and that PostFinance can still appeal to the Federal Supreme Court. That last point is press-reported and I could not verify it independently.
The case for the regulator
The strongest argument for the FDPIC's position is that a voiceprint is not a password. A customer can change a password but cannot change their voice. A leaked voiceprint is permanent, and it can be matched across any system that handles audio. The FDPIC treats voiceprints as personal data requiring special protection, and for that category the law asks for a deliberate act, not inference. Consent that people give only by failing to hang up is weak evidence of choice. On a phone line, customers cannot easily review or decline an automated notice, and a bank holds far more bargaining power than the person calling it. If the court had accepted "continuing to speak" as consent, the rule for sensitive data would have been little more than a disclosure duty.
Why the ruling is still reasonably narrow
A pro-innovation reading is available without watering down that argument. The court did not say voice authentication is unlawful. It did not ban biometrics in banking, and it did not require the deletion of every voiceprint. It set a process requirement: ask clearly, get an affirmative answer, record it. PostFinance says it has collected a spoken "yes" since early 2026, per Keystone-SDA coverage on blue News. Netzwoche reports the same change as part of a new consent process. Spoken opt-in is cheap to implement and fits naturally into a voice dialogue. That is what proportionate regulation looks like: a clear rule, a workable compliance path, and no prohibition of a security technology that many customers prefer to remembering PINs.
The extended deadline also matters. The regulator's original instruction would have forced rapid deletion of an installed base of voiceprints. A transition window until mid-February 2027 lets PostFinance re-consent customers it can reach and delete the data of those it cannot. Because the court granted that window, the outcome is closer to a negotiated compliance schedule than to a penalty.
The publication judgment is quietly important
The second ruling deserves more attention than it will get. Companies under investigation often argue that publicity is itself a sanction. The court's answer is that the FDPIC's public-information duty is statutory. Predictable publication is good for the market. Other firms running voice or face recognition can read what the regulator actually required and why. Without that, compliance guidance would come only from private legal advice. Publication also constrains the regulator, because a decision that must survive public reading is harder to stretch. The proportionality check belongs at that point, in open view, and not in unpublished settlements.
What remains unresolved
Two questions stay open. First, the court addressed consent, not necessity. Whether voice login is the least intrusive way to authenticate a caller, or whether alternatives such as one-time codes would do, is a separate proportionality question under data protection principles, and on the reported record this ruling did not decide it. Second, a spoken "yes" in a call flow raises practical questions about informed consent, such as how clearly the purpose and retention period are explained and how easily a customer can withdraw. The ruling sets the floor, and later cases will test how high it sits.
For firms outside banking, the lesson is simple. Any product that derives biometric templates from calls, video or smart devices should treat explicit opt-in as the default, document it, and plan for deletion where it is missing. For regulators, the lesson is that courts will back consent rules applied to sensitive data, but only when the remedy is targeted. A decision that fixes one clear defect and leaves room for compliance tends to survive appeal. A decision that tries to settle every question about biometrics does not.
Bottom line
Switzerland now has a judicial answer on voiceprints: express consent, not implied. The answer is demanding without being hostile to innovation. Voice authentication stays lawful, and the cost of using it is an explicit yes from the customer.