A €95 Million Phone Call, FATF's "Scam Compound" Reckoning, a Day Without Microsoft, and 672,000 Bank Customers Learn the Hard Way: Cyber's Reality Check

As of this week, the AI-voice-cloning scam that fleeced Fideuram is being treated as the textbook case, FATF has published its first landmark report on industrialised scam compounds, the ripples of the Marquis Software Solutions ransomware breach are still widening, and a global Microsoft 365 outage just sent traders back to their fax machines. Digital trust is in a trust recession, and banks are running out of denial.
If you had to pick one week to look at the state of financial cybersecurity in 2026, this might be it. The themes that security leaders have been warning about for two years, deepfakes industrialised as a service, vendor breaches cascading into dozens of banks, a single technology provider taking down half the financial world's inbox, and regulators finally finding their teeth, are no longer abstract. They are this week's headlines, and they add up to a number that should make every CISO nervous.
FATF's Landmark Report: Scam Compounds Are a Financial-System Threat
On November 10, the Financial Action Task Force (FATF) published what it called a "landmark report" calling for global action to disrupt the scam compounds that have exploded across Southeast Asia. According to FATF's own release and FF News, these compounds, often human-trafficking operations where coerced workers run romance, investment and "pig-butchering" scams around the clock, now funnel tens of billions of dollars through global financial systems each year.
The report's bite comes from the fact that it is the first time FATF has treated cyber-enabled fraud as a core AML/CFT risk rather than a sideshow. Infosecurity Magazine quoted the FATF framing: these are not isolated scams. They are a transnational industry, with crypto mixers, mule-as-a-service networks and AI deepfake tools stitched into a single supply chain. Outlook Business called it the "industrialisation of deceit."
For compliance teams, the implications are blunt. FATF will now pressure member jurisdictions to treat scam-compound flows the same way they treat terrorist financing or drug trafficking, with sanctions, dedicated FIUs, and public-private intel sharing. Any bank that has been relying on legacy transaction-monitoring rules to flag the outflows from this activity is already behind.
Fideuram: The €95 Million Voice-Cloning Scam That Changed Everything
The case study FATF would have referenced if it could is Fideuram, the private-banking arm of Intesa Sanpaolo. Over the summer, criminals used an AI-cloned voice of a senior executive, combined with a WhatsApp impersonation thread, to trick the bank into authorising overseas transfers that initially tallied €36 million and now, per Gulf News and multiple European outlets covering the wider Intesa Sanpaolo exposure, have been reported as high as €95 million once related transfers are included.
CyberNews and Itdaily both reported on the mechanics: a convincing voice note, delivered through a trusted-looking channel, in context, the perfect social-engineering cocktail. VRT reported the Italian authorities' working theory that the cloning was done with under a minute of publicly available audio.
Why Fideuram Is a Turning Point
For years, "deepfake fraud" lived in the "theoretical but rare" column of risk registers. The Fideuram loss drags it into the "material loss event" column. Trending Topics noted that this is the largest single AI-voice-cloning incident at a European bank disclosed to date, and the AML Intelligence coverage made clear that it is now being used as an industry case study inside European banking supervisors.
The British public has already drawn its own conclusion. SC World this week reported new survey data from UK banking customers expressing acute concern that their banks cannot spot AI-generated impersonation. Digidentity's parallel research, picked up by FF News, found that 81% of UK consumers believe banks should be liable for failed identity checks that lead to fraud. Translation: the public is ahead of the regulator.
DORA Starts to Bite
The EU regulator that is closest to catching up is the one born for exactly this moment. The Digital Operational Resilience Act (DORA) hit its full enforcement window in 2025, and 2026 is the year the ECB and national supervisors move from "please comply" to "here is your fine."
Mitratech's compliance teams, writing this quarter, pointed out that DORA penalties can reach 2% of total annual worldwide turnover for institutions and up to €1 million per day for ICT third-party providers deemed critical. The Recursive flagged that the first wave of enforcement actions is now being staged, not for small misses, but for the ICT third-party risk management failures that cases like Marquis Software (below) expose.
DORA is also doing something subtler: it is forcing boards to accept that cyber risk is operational-resilience risk, and must be reported with the same rigour as credit or market risk. IBM's "one year in" review called 2026 "DORA's real test", the one where supervisors actually use the powers Parliament gave them.
The Marquis Breach Keeps Widening
For a case study in exactly the ICT third-party risk DORA is designed to police, look no further than Marquis Software Solutions. What started as a single-vendor ransomware incident has, over 2026, quietly become one of the largest banking-sector breach events of the decade.
The toll this week, per American Banker: 80 banks and roughly 824,000 consumers affected, after the company disclosed that a firewall zero-day led to data exfiltration. SecurityWeek's updated figure is over 780,000 individuals directly impacted. Credit Union Times put the original disclosure at 672,075 individuals. Those are not typos, they are the growing numbers that follow any vendor-origin breach as the forensic work catches up with reality.
A Day Without Microsoft: Concentration Risk Goes Live
If Marquis shows what happens when a small vendor fails, the events of November 12–14 showed what happens when a big one does. On Thursday afternoon, Exchange Online began failing in the Americas and the problem spread worldwide, taking down or degrading Outlook, Teams, SharePoint, Copilot and, notably for security teams, Defender XDR and Intune admin dashboards. Users in at least 74 countries were affected, and Downdetector logged more than seven million reports, making it the largest outage it had recorded across any service in 2026. Microsoft reported "partial recovery" in the early hours of November 13, email queues were approaching normal later that day, and the company posted its restoration notice on the morning of November 14.
Financial markets felt it immediately. Bloomberg reported that traders were forced to do business by phone, WhatsApp and even fax. Deutsche Bank had patchy access to email, Euronext, Nasdaq and Deutsche Boerse reported email problems, traders in Singapore said some banks could not complete financing deals, and Japanese securities firms reverted to fax to confirm trades. Trade Finance Global described banks keeping transactions moving by phone as the outage stretched into a second day. Anderson Economic Group has put the economic cost at around $7 billion.
There is an uncomfortable irony here for anyone who read the Fideuram section above: when the corporate channel went dark, the industry's fallback was the same consumer messaging app that the voice-cloning fraudsters used to make their impersonation look legitimate.
What Went Wrong
From the outset, Microsoft said there was no indication of malicious activity, this was not a cyberattack. Its preliminary post-incident report, dated November 15, describes a classic cascading failure rather than a single mistake:
A defect in a fix deployed to Microsoft 365 networking gateway servers from July, a lock-contention issue in an internal Windows process that, under certain traffic conditions, exhausted processor and memory. It ran globally for more than three months without visible problems, and testing missed it because it focused on less powerful machines; the bug was worse on newer, many-core hardware.
About fifteen hours before the outage, Teams traffic was moved onto the primary gateway shared with Exchange Online.
Roughly 14% of gateway capacity was already offline for an unrelated reason.
Automated recovery then moved connections to healthy servers, triggering the same fault there, and a separate failure in fault-containment boundaries allowed the problem to spread globally. At the worst point, more than 80% of gateway servers were unresponsive.
Microsoft's fix was to roll Windows connection handling back to its earlier state, cycle large groups of servers out of service and restore traffic gradually, while extending email delivery windows so queued messages were not bounced. Repair items are due in the final report on November 20.
Why It Matters for Banks
Microsoft is not just any vendor. In November 2025, the European Supervisory Authorities included it on their first list of 19 critical ICT third-party providers under DORA, and since July 2026 the Bank of England, PRA and FCA have been overseeing Microsoft as a designated Critical Third Party in the UK, alongside AWS, Google Cloud and Oracle. That designation does not mean Microsoft breached DORA, but legal commentators note that the outage could meet DORA's major-incident reporting thresholds for some affected financial entities, which puts the spotlight back on the institutions' own continuity plans. In the UK, the FCA monitored the impact on financial firms and advisers were quick to point to its calls for firms to review their resilience, while minister Luke Pollard urged "greater resilience" against technology failures.
The Pattern
The Marquis story and the Microsoft outage, like the parallel reports from Sophos' State of Ransomware in Financial Services 2026, reveal the pattern that DORA, FATF and the ECB are all converging on: your biggest cyber risk is not your own perimeter. Sometimes it's the smaller software vendor three layers down in your supply chain who runs an exposed appliance and reuses credentials. Sometimes it's the hyperscaler at the very top of it, where a three-month-old bug can switch off email, collaboration and security tooling for an entire industry at once, no attacker required.
Black Kite's 2026 Financial Services Report, covered this week by Unite.AI, put hard numbers on it: a growing majority of material financial-services cyber incidents in 2026 originated at a third-party vendor, not at the regulated institution itself. That is exactly the hole DORA was written to close.
eIDAS 2.0: The Digital Identity Wallet Starts to Matter
Against this backdrop, the European Digital Identity Wallet, delivered under eIDAS 2.0, is suddenly less of a regulatory abstraction and more of a possible cure. Deloitte Luxembourg described it this week as a "revolution reshaping financial services"; Worldline's banking team argued that if adopted at scale, EUDI Wallet can slash onboarding fraud and account-takeover risk by replacing passwords and knowledge-based checks with cryptographic identity assertions bound to a device the user controls.
The wallet obligation for banks moves meaningfully into 2027, per Deloitte and Capco, but the time to invest is now. Namirial's team flagged the "decide in 2026" question succinctly: do you integrate the wallet into onboarding and strong customer authentication early, and get the fraud-reduction dividend, or do you wait, and meet the deadline with the minimum effort?
For the banks currently losing money to voice-cloning scams, that question answers itself.
The Takeaway
The headlines this week don't tell a story about new attack vectors. They tell a story about industrialisation, and concentration. AI deepfakes are no longer artisan work, they are a service. Ransomware groups no longer target one bank, they target the vendor who services eighty. Scam compounds no longer defraud one victim at a time, they run at compound scale, often with state-level tolerance. And when one cloud provider stumbles, it doesn't take down a bank, it takes down the market's ability to send an email.
The regulatory response is finally catching up. FATF is treating cyber-enabled fraud as the AML priority it is. DORA is moving from publication to enforcement, and the EU and UK now have direct oversight of the hyperscalers that underpin the sector. eIDAS 2.0 is handing banks the identity rail they need. And the consumer survey data from the UK is a warning: trust in bank defences is leaking, and banks will be judged more on their response than on the sophistication of the attacks they stop.
For CISOs and chief risk officers, the triage list for the rest of 2026 writes itself. Third-party and ICT risk, audited to DORA standards, from the smallest vendor to the largest cloud provider. Tested fallback channels for when the primary communications stack goes dark, secured so they don't become the next impersonation vector. AI-impersonation defence at the call-centre and executive-approval layer. Customer-facing identity assurance tied to a wallet, not a password. Everything else is a nice-to-have.



