On the morning of Tuesday 20 October 2026, customers of the digital bank Revolut in the United Kingdom and Ireland opened their applications to find card payments and ATM withdrawals unavailable or only partially working. According to Reuters, the disruption traces back to the London operations of Google’s cloud infrastructure, the europe-west2 data centre, whose engineering teams said they were investigating the issue. Revolut stated on X that it was experiencing a service disruption affecting multiple services, while confirming that all customer money and data were safe. The same failure affected other institutions that morning, with ING Germany reporting problems logging into online banking and its application, and British banks including Halifax, Lloyds and HSBC fielding customer complaints about online services and applications.
For readers living in France with an account at a digital bank licensed elsewhere in Europe — expatriates, international students, cross-border workers and investors — the episode raises two precise legal questions. The first concerns your money and the payments that fail while the screens stay dark: refunds, fees, default interest and the protection of deposits. The second concerns responsibility when the bank points to its cloud provider: French contract and payment law answers that question without requiring customers to pursue a data centre operator. This article sets out that answer, provision by provision, starting with what the morning of 20 October changed and what it did not change.
I. Your money and your payments during the blackout: what French law still guarantees
A. Failed, delayed or duplicated payments: the bank refunds, fees are capped and interest runs
The first reflex when a payment fails during a general outage is to assume that an exceptional event suspends every right. French payment law takes the opposite starting point. When the payment order is given by the payer, French law provides that his payment service provider is, subject to stated reservations, liable for proper execution of the payment transaction toward the payer, in the words of the statute, “responsable de la bonne exécution de l’opération de paiement à l’égard du payeur”. Where the payer’s provider is liable for a badly executed transaction, the same provision requires it to return the amount without delay: “il restitue sans tarder son montant au payeur”. The rule is Article L. 133-22 of the French Monetary and Financial Code, and it applies regardless of whether the technical cause sits inside the bank or at one of its subcontractors, because the customer-facing obligation is the bank’s own.
The protection is even stricter for transactions the customer never authorised. Article L. 133-18 of the same code provides that the payer’s provider refunds the amount of the unauthorised transaction immediately after learning of it or being informed of it, and in any event no later than the end of the first business day following, in the statutory words, “le prestataire de services de paiement du payeur rembourse au payeur le montant de l’opération non autorisée immédiatement après avoir pris connaissance de l’opération ou après en avoir été informé, et en tout état de cause au plus tard à la fin du premier jour ouvrable suivant”. The same article attaches penalty interest to any failure to refund on time: “Les sommes dues produisent intérêt au taux légal majoré de cinq points”, rising to the legal rate increased by ten points beyond seven days of delay, since “Au-delà de sept jours de retard, les sommes dues produisent intérêt au taux légal majoré de dix points”, and by fifteen points beyond thirty days, since “Au-delà de trente jours de retard, les sommes dues produisent intérêt au taux légal majoré de quinze points”. An outage that delays refunds therefore becomes more expensive for the provider every week it lasts.
These rights come with a deadline that customers should treat as absolute. Article L. 133-24 of the Monetary and Financial Code requires the payment service user to report an unauthorised or badly executed transaction without delay and no later than thirteen months after the debit date, on pain of foreclosure: “L’utilisateur de services de paiement signale, sans tarder, à son prestataire de services de paiement une opération de paiement non autorisée ou mal exécutée et au plus tard dans les treize mois suivant la date de débit sous peine de forclusion”. The thirteen-month period runs from the debit date, and the only statutory exception is where the provider failed to supply the transaction information it owed. Concretely, anyone whose card was debited twice when systems recovered, whose transfer never arrived, or whose direct debit bounced during the outage should notify the bank in writing now, keep proof of the message, and diarise the thirteen-month limit from each debit date.
Fees charged in connection with the incident are subject to their own ceiling. Article L. 133-26 of the same code provides that the amount of fees following a payment incident, other than a cheque rejection, is capped under conditions set by decree according to the nature and amount of the incident, and never exceeds the amount of the incident itself: “Le montant des frais consécutifs à un incident de paiement autre que le rejet d’un chèque est plafonné dans des conditions fixées par décret, en fonction de la nature et du montant de l’incident, sans excéder en tout état de cause ce dernier montant”. Intervention fees stacked on top of each other after repeated rejections of the same transaction must therefore be examined line by line, and the portion exceeding the first rejection must be refunded, since the same article requires the provider to return fees collected on repeated rejections of the same payment beyond the amount taken for the first rejection.
The courts apply these provisions without indulgence toward the provider. On 22 January 2020 the Commercial Chamber of the Court of Cassation, in case number 18-18.640, dealt with a bank that had been ordered to refund contested payment transactions, and the decision records that the court of appeal had acted “après avoir condamné la banque à rembourser à celle-ci le montant des opérations contestées et des frais de commissions et d’interventions”, that is, after ordering the bank to refund the customer the amount of the contested transactions together with commission and intervention fees. The part of that reasoning that matters for an outage is structural: the refund covers the principal and the fee layer built on top of the incident. A year earlier, on 26 June 2019 in case number 18-12.581, the same chamber rejected a bank’s appeal against an order to refund a customer, stating simply “REJETTE le pourvoi”. The judgment under appeal in that case had ordered the bank to pay the customer the sum of 2,753.56 euros with interest at the legal rate running from the formal demand of 12 June 2014. Two lessons follow for the customers affected on 20 October 2026: the provider, not the customer, carries the burden of sorting out what happened in its systems, and once liability is established the meter runs from the formal demand, at the legal rate, without waiting for the technical post-mortem.
One boundary must be stated clearly, because outages generate exactly the kind of sprawling loss schedules that courts trim back. Consequential losses — a missed train, a hotel night paid twice, a business appointment lost — are recoverable only within the filter of Article 1231-4 of the Civil Code, under which damages include only what is an immediate and direct consequence of non-performance: “les dommages et intérêts ne comprennent que ce qui est une suite immédiate et directe de l’inexécution”. Customers should therefore document every loss with receipts and timestamps, while understanding that the refund of the payment itself stands on far firmer ground than the remote consequences of having been unable to pay for a few hours.
B. An outage is not a loss: deposits remain protected and the European passport designates the guarantor
The most reassuring statement of the day came from Revolut itself, which confirmed on X that customer money and data were safe while services were disrupted. Legally, that statement reflects a structural feature of banking law rather than a commercial gesture: a temporary inability to access funds is not a loss of funds, and two separate bodies of rules protect the two distinct risks. The first risk — the bank no longer being able to return deposits at all — is covered by the deposit-guarantee mechanism. The second risk — being unable to reach your money for hours — is a service failure governed by the liability rules examined above, not by the guarantee.
For banks licensed in France, the guarantee mechanism is organised by Articles L. 312-4 and following of the Monetary and Financial Code. Credit institutions licensed in France join the deposit guarantee and resolution fund, which by statute has the mission of managing and implementing the deposit-guarantee mechanism, in the words of Article L. 312-4: “Le fonds de garantie des dépôts et de résolution a pour mission de gérer et de mettre en œuvre : 1° Le mécanisme de garantie des dépôts”. Article L. 312-4-1 defines the scope of the cover: “La garantie des dépôts couvre, dans la limite d’un plafond, les fonds laissés en compte auprès d’un établissement de crédit et libellés en euros ou dans la devise d’un autre Etat”, expressly including sums corresponding to payment transactions in progress carried out for an identified person in the course of normal banking operations. The trigger is set by Article L. 312-5: “Le mécanisme de garantie des dépôts est mis en œuvre sur demande de l’Autorité de contrôle prudentiel et de résolution lorsque celle-ci constate qu’un établissement de crédit n’est plus en mesure de restituer, immédiatement ou à terme rapproché, les fonds mentionnés au I de l’article L. 312-4-1 , et au plus tard cinq jours ouvrables après avoir établi pour la première fois que cet établissement de crédit ne les a pas restitués.” Nothing in the events of 20 October 2026 resembles that trigger: the institutions concerned never suggested they were unable to return deposits, and a cloud outage measured in hours, with money confirmed safe, sits on the service-failure side of the line, not on the guarantee side.
Customers in France who bank with an institution licensed in another European Union member state fall under a parallel but distinct arrangement. French law expressly allows it: within the limit of the services it is authorised to provide in its home member state, any credit institution may operate on French territory through branches and under the freedom to provide services, as Article L. 511-22 of the Monetary and Financial Code states, “tout établissement de crédit peut, sur le territoire de la France métropolitaine, des départements d’outre-mer, du Département-Région de Mayotte et de Saint-Martin, établir des succursales pour fournir des services bancaires et intervenir en libre prestation de services dans les conditions définies à l’article L. 511-24”. Under the European passport, deposits held with such an institution are covered by the home state’s guarantee scheme, not by the French fund. This is precisely the situation of French residents whose accounts are held with Revolut Bank UAB: the bank’s own depositor information for France states that deposits with Revolut Bank UAB are insured by the Lithuanian public institution for deposit and investment insurance, up to 100,000 euros per depositor for all deposits held with that bank. The practical consequence is a two-track reflex. As long as the institution confirms that funds are safe, the remedy for blocked access runs through the liability and refund rules, not through any guarantee fund. Only if an institution were found unable to return deposits would the guarantee of the licensing state be activated at the supervisor’s request.
A comparison helps to keep the categories straight. Readers of this publication will recall the very different situation of funds frozen under European Union sanctions, where named persons and entities cannot access their assets by operation of law and must seek a Treasury authorisation or challenge the listing before the EU courts, as examined in the analysis of the October 2026 record sanctions listings. A sanctions freeze is a legal prohibition addressed to the holder; a cloud outage is a technical failure addressed to nobody. The first requires an authorisation to move funds; the second requires the provider to restore service and to make the customer whole for the incident.
II. The cloud behind the bank: who bears the outage risk when a provider fails
A. Your contract is with the bank, not with its data centre: the force majeure test applied to an outsourced outage
When service returns, the predictable defence is that the failure came from upstream — from Google Cloud’s europe-west2 data centre rather than from the bank’s own systems — and that nobody could have prevented it. French contract law channels that defence into a narrow gate with three cumulative locks. Article 1218 of the Civil Code defines force majeure in contract as an event beyond the debtor’s control, which could not reasonably have been foreseen when the contract was concluded and whose effects cannot be avoided by appropriate measures, preventing the debtor from performing his obligation: “Il y a force majeure en matière contractuelle lorsqu’un événement échappant au contrôle du débiteur, qui ne pouvait être raisonnablement prévu lors de la conclusion du contrat et dont les effets ne peuvent être évités par des mesures appropriées, empêche l’exécution de son obligation par le débiteur.” The same article distinguishes temporary from definitive prevention: “Si l’empêchement est temporaire, l’exécution de l’obligation est suspendue à moins que le retard qui en résulterait ne justifie la résolution du contrat.” An outage lasting hours is the textbook case of a temporary impediment: performance is suspended, not extinguished, and the obligations revive with the systems.
The sanction for failing that test is stated by Article 1231-1 of the Civil Code: “Le débiteur est condamné, s’il y a lieu, au paiement de dommages et intérêts soit à raison de l’inexécution de l’obligation, soit à raison du retard dans l’exécution, s’il ne justifie pas que l’exécution a été empêchée par la force majeure.” Applied to 20 October 2026, this allocation matters in two ways. First, the debtor of the account-holding and payment obligations is the bank, and it is the bank that must justify force majeure; the customer has no contract with the cloud operator and no duty to pursue it. Second, each of the three conditions can be tested against the facts of an outsourced infrastructure. An incident at a subcontractor chosen, configured and paid by the debtor sits uneasily with the requirement of an event beyond the debtor’s control. A regional failure of a single data centre sits uneasily with unforeseeability for a financial institution whose business is continuous availability, and even more uneasily with unavoidability where replication across regions and providers belongs to the standard toolkit of the sector. None of this decides any individual dispute in advance — only a court ruling on fully debated evidence could do that — but it explains why an upstream outage is the beginning of the bank’s explanation rather than the end of the customer’s claim.
The same prudence applies in reverse. Customers sometimes assume that every consequence of the outage is automatically chargeable to the bank, from a missed connecting flight to a lost business deal. The immediate-and-direct filter of Article 1231-4 recalled above trims those schedules back to what flows directly from the non-performance, and courts examine causation document by document. The strongest files will therefore pair each claimed loss with its receipt, its timestamp and the proof that no alternative means of payment was reasonably available at the critical moment — for example a second card on a different network, which many travellers carry precisely for this reason. Keeping that second means of payment is not a legal duty, but its absence or presence will inevitably colour any debate about avoidability of the loss.
B. Supervisors now watch the cloud itself: critical providers, incident reporting and what oversight changes for customers
The second institutional lesson of 20 October 2026 is that regulators on both sides of the Channel already treat large cloud operators as part of the financial system. According to Reuters, the United Kingdom in July 2026 designated four technology companies, including Google Cloud EMEA Limited, as critical third-party suppliers for its financial sector, bringing them under direct regulatory oversight. The same reporting recalls that European Union regulators had designated nineteen technology companies the previous year, including Google Cloud, as critical third-party computing providers on which the bloc’s finance industry depends. In other words, the concentration risk revealed on Tuesday morning — one London data centre affecting banks in several countries — is exactly the risk supervisors say they are now mapping, measuring and constraining.
For customers, this supervisory layer changes the dispute landscape in three practical ways. First, financial institutions must now detect, classify and report major technology incidents to their supervisors within tight deadlines, which means the outage of 20 October 2026 will generate an official paper trail — incident reports, root-cause analyses and remediation plans — that did not exist in the era when an IT failure was purely an internal matter. A customer who files a written complaint in the days following the incident therefore writes into a file that the supervisor may later read. Second, outsourcing no longer dilutes responsibility in the supervisor’s eyes: a bank remains answerable for the services it entrusts to its providers, and concentration on a single cloud region is precisely the kind of arrangement supervisors question. Third, the designation of critical providers opens a direct oversight channel over the largest technology suppliers themselves, so that the resilience of a data centre serving millions of banking customers is examined before failures rather than only after them.
In France, the authority holding these threads together is the Autorité de contrôle prudentiel et de résolution. Article L. 612-1 of the Monetary and Financial Code entrusts it with an independent watch over the stability of the financial system and the protection of customers, stating that “L’Autorité de contrôle prudentiel et de résolution veille, de manière indépendante, à la préservation de la stabilité du système financier et à la protection des clients, assurés, adhérents et bénéficiaires des personnes soumises à son contrôle.” The same provision charges it with verifying compliance with directly applicable European rules and with the provisions of the Monetary and Financial Code: “L’Autorité contrôle le respect par ces personnes des dispositions européennes qui leur sont directement applicables, des dispositions du code monétaire et financier ainsi que des dispositions réglementaires prévues pour son application”. For a customer of an institution operating in France under the European passport, the competent supervisor for prudential matters remains in principle the authority of the licensing state, but the French authority’s client-protection mission and its control over compliance with applicable European provisions give complaints filed from France a genuine institutional addressee. The recommended sequence stays simple: written complaint to the bank with statements and timestamps, escalation through the dispute-resolution scheme named in the account terms if the answer is unsatisfactory, and a parallel report to the competent supervisor so the incident enters the official record.
Conclusion
The outage of 20 October 2026 changed less than the dark screens suggested. Deposits were never lost: they were temporarily unreachable, which banking law treats as a service failure rather than a guarantee event, with the French fund standing behind French-licensed institutions and the home-state scheme — the Lithuanian guarantee up to 100,000 euros per depositor for Revolut Bank UAB customers — behind passported ones. Failed or duplicated payments trigger the provider’s liability for proper execution, refunds by the end of the first business day for unauthorised transactions with penalty interest running at five, ten and then fifteen points above the legal rate, capped incident fees, and a thirteen-month deadline from each debit date that customers should secure in writing now. And when the bank gestures toward its cloud provider, French contract law answers that the debtor is the bank, that force majeure requires an unforeseeable, irresistible and unavoidable event, and that a temporary regional failure of chosen infrastructure will struggle to pass that triple test — while remote consequential losses still have to clear the immediate-and-direct filter. Behind the dispute stands a supervisory shift both London and Brussels have already made: the cloud operators on which banks run are now regulated as critical providers themselves. File the complaint, keep every timestamp, and let the paper trail do the rest.