Skip to content

// analysis

Deutsche Bank and Epstein: the compliance exception in the internal record

Open the documentsThe source documents

The documents cited in the analysis and, where identified, their associated passages. An automatically detected association is not a verified claim.

  1. initial 9 August repository version date source non détectée · Automatically detected association, not reviewed
    Its SHA-256 digest matches the file in the initial 9 August repository version : 62c3bc2acb77130b23c93d3704eae923027423804e93f30988eb75f840fe7eae The server-side push and build history provides dated evidence.
Illustration for the analysis: Deutsche Bank and Epstein: the compliance exception in the internal record
Sources and reading context

Topics

dated revision: November 18, 2026primary sourcesno tracker
Contents8 sections

The Inside Deutsche Bank investigation, published on 17 November 2026, puts a larger figure and new internal records behind the bank’s relationship with Jeffrey Epstein. The consortium reports $581.9 million in transactions subsequently identified as suspicious. Its accounts of internal exchanges follow the permissions that allowed a known risk to become an enduring client. S06

l0g addressed that mechanism in our analysis published on 9 August, included in the Epstein’s Money collection. The French article and EPUB existed 100 days before the consortium’s first publications. Their central argument was already explicit: a succession of approvals and compromised controls had made an exceptional client increasingly ordinary inside the institution. S01 S14 S15

The groundwork came from the New York regulator’s 2020 findings, Reuters’ investigation of the protracted exit in February 2026 and Senator Ron Wyden’s August banking report. These sources made our earlier analysis possible. November brings the work of an international consortium with access to internal material. The new findings belong to those investigators, and their evidence provides a more detailed basis for testing our earlier argument. S02 S04 S05

This analysis is current to 18 November 2026. We have examined the public investigations and official documents cited below. Access to the underlying leak remains with the consortium and its research infrastructure. S09 S10

A known risk becomes reusable permission

Commercial ambition and a known risk shared the same acquisition file. In April 2013, it paired Epstein’s criminal history with prospective annual revenue of $2 million to $4 million. The decisive May email invoked conversations with legal and anti-money-laundering officials. The NYDFS found no separate record of those discussions or initial meeting of the regional reputational-risk committee. S02

Anti-money-laundering controls, usually abbreviated to AML, help banks identify and report potentially unlawful activity. Know-your-customer checks, or KYC, establish who the client is and the expected purpose of the relationship. A high-risk designation should affect the scrutiny applied to actual transactions. Its value depends on the decisions it changes.

In our reading, that approval became a substitute for asking whether particular activities were justified. A senior commercial officer had represented that the relationship could proceed. Staff taking over the client could refer back to his email, treating the sensitive issue as settled. The organisation had produced an answer that could travel more easily than its underlying reasoning.

Responsibility follows the surviving record. The email establishes the commercial executive’s account of a conversation, with the legal and AML officials’ agreement reported through him. A directly recorded approval by those officials would be a different piece of evidence. Keeping that distinction visible allows the reconstruction to rest on the documents actually available. S02

The NYDFS described conditions imposed after a 2015 review that failed to reach all the relevant staff. Transaction monitoring also came to assess unusual activity against Epstein’s own previous activity. In its 2020 findings, the regulator identified a 2017 alert closed on that basis. S02

Repeated exceptions can gradually acquire the appearance of normal behaviour. Each earlier approval supplies context for the next alert, making a transaction less surprising because something similar has happened before. The original concern may remain unresolved throughout. Reassurance then comes from a history of activity the bank itself has permitted.

This is especially dangerous when the concern lies in the purpose of a payment. Comparing transactions with past behaviour helps to detect sudden changes. A stable practice poses a harder problem: repetition can look reassuring even when that practice fits the risk already identified. Monitoring has to keep the original concern in view as the history of authorised activity grows.

How the control chain changed Mechanism established by the NYDFS: known risk, approval asserted in an email, monitoring conditions poorly communicated, and activity then assessed against the client’s own history. The dates cover 2013 to 2017. How the control chain changedDeutsche Bank · 2013–2017 · documented stepsKnown riskApril 2013Approval assertedemail · May 2013Conditions setEarly 2015Alert closedMarch 2017CommunicationClient’s own historybecomes the monitoring benchmark The control chain2013 → 2017 · NYDFS, July 2020Known riskApril 2013Approval citedby emailConditionspoorly relayedAlert closedclient’s own historyPast activity becomes the benchmark.
Source: NYDFS order published in July 2020, §§ 19–22 and 31–42. Decisions and implementation during 2013–2017. The approval asserted in May 2013 appears in a front-office email. S02

A report produced by independent monitors in 2020, described by the consortium, identified extensions of the relationship without appropriate control documentation. Services had expanded while the record of the relevant checks failed to keep pace. Earlier permission could continue to support a growing relationship, even as each extension called for an assessment of the additional risk. S06

The larger total changes the measurement problem

The consortium attributes its $581.9 million figure to an internal US Financial Crimes Intelligence Unit document covering 21 suspicious activity reports and approximately 3,000 transactions between October 2013 and May 2019. A suspicious activity report, or SAR, alerts authorities to activity a financial institution considers suspicious. S06

The August Wyden report described more than $250 million in wire transfers retrospectively flagged by Deutsche Bank. Its account included an example of 1,140 wires worth $147 million. Those were findings attributed to investigators who had reviewed records at the Treasury. S05

Reconciling these totals requires the underlying populations of payments and their selection rules. The earlier figure was a floor. The new aggregate expands the public picture, but the apparent gap also depends on which transactions were included and how they were counted.

A bank’s financial-intelligence record is organised around reporting obligations. Reconstructing the economics of a network means tracing the money from its entry to its eventual use or retention. A collection of suspicious transactions may contain incoming payments, outgoing payments and successive transfers of funds already counted earlier in their journey. Those movements have to be connected before the total can describe the resources available to the account holder.

FinCEN’s guidance on completing SARs discusses the treatment of flows and the calculation of reported amounts. It reinforces the need to read the reporting method before turning an aggregate into an economic measure. S16

Suppose the same funds move from one account to another and then to a third recipient. Each step may matter to an investigator. Counting the steps describes movement through the system. Identifying the original resources available to the account holder requires a reconstruction of those links. The two exercises can both be useful while producing different totals.

Read as the value of transactions covered by the suspicious-activity reporting described by the consortium, the November figure has a defined purpose. It measures reported movements. A fortune is a stock of assets at a particular date, while criminal proceeds require evidence about the origin of the funds. Preserving those distinctions gives the larger figure its analytical value.

Our earlier cross-bank investigation made this accounting discipline central to the dossier. Different institutions can observe separate legs of the same transfer. A larger number becomes analytically useful when readers understand the perimeter, rather than simply the scale.

Timing changes the usefulness of that record. A retrospective investigation can recover a valuable trail for authorities, sometimes years after the money moved. Preventive monitoring has to work while the relationship and its payments are still active. The Wyden report’s criticism concerns that delay. Reconstructing the history reveals the distance between a risk that could have been recognised and the bank’s eventual response. S05

An exit that kept supplying reassurance

The distinction between deciding to leave and completing the exit was already part of our August analysis. Deutsche Bank notified Epstein on 21 December 2018 that it would stop servicing his accounts. Reuters subsequently documented activity after the transfer deadline, including a reference letter in March 2019 and accounts still holding balances in May. S02 S04

A bank’s exit can involve settling securities positions, moving cash and finding another institution willing to accept the assets. Someone must control the instructions the departing client can still give during that process. Describing the relationship as “terminated” can conceal this interval, allowing the date of a decision to acquire the appearance of a complete operational break.

Following the services gives a more useful chronology. It identifies the transfers still permitted, the staff with authority to approve them and the point at which each account stopped functioning. Until those permissions have been withdrawn, the exit remains an active exercise in managing the client’s risk. An announcement supplies its intended direction, while the available services show how far the bank has actually travelled.

The reference letters belong to that operational history. The NYDFS had already recorded letters issued to two institutions despite the offboarding decision. Reuters added reporting on their timing. The November investigations give the episode a clearer internal context. S02 S04

According to the consortium, a March 2020 email copied to senior executives acknowledged the lack of a policy for these references. Der Standard also reports that the employee writing them knew about problematic activity and had participated in the exit decision. S06 S08

Alongside the assets, information was moving to another institution that could use it to decide whether to accept the client. A letter carries the authority of the bank’s name while compressing a complicated relationship into a short statement about account operation. Its wording can influence the confidence with which the recipient approaches the incoming business.

An orderly transfer may suit the commercial relationship manager who knows the client. For the receiving bank, the value of a reference depends on its accuracy in light of the risks known to the sender. An exit procedure must determine who can speak for the institution and how that person’s assurances are checked against the reasons for the departure.

Where the sender’s own concern helped trigger the departure, reassurance warrants particular scrutiny. The assets are leaving because the relationship has become unacceptable, yet the accompanying language may help another bank to welcome it. Reconstructing the wording and approval of the reference follows the risk across the boundary between the two institutions.

Der Standard dates the closure of Epstein’s last accounts to 9 July 2019, following his arrest. The ZDF account similarly places final closure after the arrest. This extends the operational history beyond the December decision. S07 S08

Two hundred days from the decision to the final closures Proportional timeline from 21 December 2018 to 9 July 2019, spanning 200 days. Deadline on 28 February, reference letter on 18 March, arrest on 6 July and the final two accounts closed on 9 July. The 131-day interval from 28 February to 9 July does not represent continuous banking activity. The bank’s exit spans 200 daysEvent dates · scale in calendar days21 Dec 2018Exit decision28 Feb 2019Deadline18 MarchReference letter6 JulyArrest9 JulyFinal 2 accounts131 days after the deadline An exit spanning 200 days21 Dec 2018 → 9 July 201921 DecDecision28 FebDeadline18 MarchReference9 JulyFinal 2 accounts6 JulyArrest131 days after the deadline
Sources: Reuters and Inside Deutsche Bank. Proportional calendar-day scale: 200 days from 21 December 2018 to 9 July 2019, including 131 after the 28 February deadline. Points mark documented events. S04 · S08

This is a concrete point of continuity with our August analysis, which asked who had approved the references and the services maintained during the exit. The new account supplies part of the answer by describing the absence of a policy. Establishing the full chain still requires the preparation and approval of the letters, together with the information supplied to their recipients.

Germany’s public record has also moved on

Germany’s supervisory record also requires an update. Our August account described the government’s public position then available. Parliamentary replies published before the leak have since clarified which American documents BaFin examined and the limits of the information released to the public.

In replies dated 5 October and published on 9 October, the German government stated that BaFin had examined both the NYDFS order and the Wyden report. It emphasised the US authorities’ responsibility for the American banking relationship and referred to earlier answers, including classified material. A question about the joint BaFin and financial-intelligence task force also received a partly classified answer. S12

Part of the supervisory response therefore exists within protected parliamentary procedures. Public readers have the government’s confirmation that the documents were examined and its account of the competent authorities. Access to the substance of some supervisory work remains restricted, shaping how far outside scrutiny can follow it.

Another answer, published on 21 September and summarised by parliament on 12 October, clarifies two requests sent earlier in 2026. They concerned possible Epstein connections of significant owners of German banks and their reliability under ownership-control rules. The government said the information reviewed provided no indication that those owners were unreliable. The scope concerns bank ownership and must be distinguished from a review of the handling of Epstein as a customer. S13

An assessment of a bank owner and a review of customer monitoring address different conduct. A criminal investigation has its own evidentiary requirements. Following each institution’s mandate makes it possible to identify where findings overlap and which authority can act on a particular failure.

Publication and regulatory access also have separate chronologies. Establishing when an authority received a particular record would help to assess the scope of its earlier work and any response to the November reporting. The competent institutions hold part of that history. Further disclosure could reveal how the evidence travelled from the bank to its supervisors.

One hundred days, with a preserved text

On 9 August 2026, l0g published an analysis of the compliance-exception mechanism and included it in the French EPUB. The consortium began publishing findings based on internal records on 17 November. 100 days separate those dates. The earlier text provides a fixed point against which to compare the new evidence. S01 S14 S15

The earlier article linked the known risk to the approval chain, the handling of alerts and the drawn-out exit. Its questions included the authority behind services maintained during offboarding and the references sent to other banks. The original conclusion was explicit: ‘Controls can know the risk, label it and still learn to live with it.’ Readers can inspect that argument in the preserved text. S01

The integrity check is useful because a date displayed on a web page can be edited. A historical repository record preserves the earlier text. Comparing the initial EPUB’s bytes with the file publicly available today adds a check on whether the edition has changed. The evidence supports a specific text and file at a specific point in the publication history. S15

Linking an unchanged file to a dated record Two copies of the French EPUB share an identical SHA-256 digest: the file recorded on GitHub on 9 August 2026 and the public copy checked on 18 November. A second axis shows the 100 days between the 9 August record and Inside Deutsche Bank on 17 November. The Sigstore attestation covers the complete site archive, rather than a standalone signature on this EPUB. Linking an unchanged file to a dated recordFrench EPUB · file integrity and repository chronologyIdentical SHA-25662c3bc2a…40fe7eae9 August 2026GitHub record18 Nov 2026Public copy checked100 days9 Aug · record17 Nov · investigation9 August release: Sigstore attestation step succeeded An unchanged, dated fileFrench EPUB · checked on 18 NovemberIdentical SHA-25662c3bc2a…40fe7eae9 August18 Nov100 days9 Aug · record17 Nov · investigation9 August · Sigstore step succeeded
Checked on 18 November 2026: the files have the same digest. The 100 days link the 9 August repository record to publication on 17 November. The historical run records the release attestation. S14 · S15

The preserved text establishes the scope of our earlier analysis: it had identified a mechanism subsequently illustrated by additional internal evidence. The underlying regulatory findings were already public, and the consortium’s investigation was underway before publication. Credit for the original findings belongs to the regulator and investigators whose work supplied the evidence, including the consortium’s new reporting. S02 S10

That surviving argument now meets a more detailed internal record. Its value lies in how well it explains the relationship as the evidence grows. The new material on control documentation and exit references develops the same institutional problem, while supplying specific episodes that deserve closer reconstruction.

Reform has to change the decision

Deutsche Bank’s 17 November statement says the matters reported are historical and points to €7 billion of direct investment in its control environment between 2019 and 2025. It describes a transformed institution with stronger risk management and acknowledges continuing remediation. These are the bank’s own statements about its reform programme. S11

Spending measures the resources committed to controls. Their effectiveness depends on how people use them. Following an actual case from the first concern to the resulting action would show whether staff received a restriction they could apply and whether it changed the services available to the client. That is where an investment programme reaches the commercial relationship.

The ZDF reports that independent reviewers feared the customer base might contain other cases comparable in scale to Epstein. That warning concerns the possibility of a broader control weakness. It gives a reason to test the system across customers, with the findings and the review’s limits kept visible. S07

Independent monitoring already formed part of the NYDFS order’s response to the failures it described. The order also distinguished that administrative settlement from the responsibilities of other authorities. Assessing subsequent reform means following what happened under those arrangements, including the findings that returned to supervisors and the changes made in response. S02 S03

Our interpretation of the accumulated record is that an exception becomes durable when each part of the institution can point to an earlier decision as sufficient authority to proceed. The challenge of reform is to make a new concern capable of reopening that decision. A customer file can contain the right facts for years while the practical permissions around it keep expanding.

The internal accounts add weight and specificity to the problem l0g analysed in August, especially the production of reassurance during an exit. Further documents could trace who had authority to stop the activity and how the relevant warnings reached them. The result would be a closer account of the point at which knowledge of the risk finally changed the services the client could obtain.

Further reading: Epstein’s Money and the separate Barclays, Staley and Maxwell investigation.

Sources

  1. S01 · l0g · Published: 2026-08-09 · Accessed: 2026-11-18 Deutsche Bank and Epstein: the compliance exception
    Passages and scope

    Original analysis: email approval, transmission of conditions, monitoring and exit. Compared with the text in commit e7c35851ae11247b737888f829b2f952e43c5ae7.

  2. S02 · NYDFS · Published: 2020-07-07 · Accessed: 2026-11-18 Consent order concerning Deutsche Bank
    Passages and scope

    Signed on 6 July and published on 7 July 2020. Paragraphs 19–25: onboarding and reported approval. Paragraphs 31–42: conditions and monitoring. Paragraphs 53–54: exit and references. Paragraphs 7 and 114: penalty covering Epstein, FBME and Danske Estonia. Administrative settlement.

  3. S03 · NYDFS · Published: 2020-07-07 · Accessed: 2026-11-18 Compliance failures and the $150 million penalty
    Passages and scope

    Regulator’s release: procedural failures, transmission and interpretation of conditions, scope of the penalty. Read with the full order.

  4. S04 · Reuters · Published: 2026-02-11 · Accessed: 2026-11-18 Reuters investigation into the prolonged Epstein exit
    Passages and scope

    Exit chronology based on DOJ documents examined by Reuters: 21 December 2018 letter, 28 February 2019 deadline, 18 March reference and services during the exit. Records may be incomplete.

  5. S05 · U.S. Senate Finance Committee · Ron Wyden’s staff · Published: 2026-08-04 · Accessed: 2026-11-18 Looking the Other Way: report on banks and Epstein
    Passages and scope

    Cover dated 4 August 2026 despite the November URL path. Section X, PDF pages 44–45: more than $250 million and an example covering 1,140 wires worth $147 million. Senate investigators’ findings and recommendations, drawing on an in-camera Treasury review.

  6. S06 · Inside Deutsche Bank · Newslaundry · Published: 2026-11-17 · Accessed: 2026-11-18 Consortium investigation into Epstein-linked transactions
    Passages and scope

    Khadija Sharife, Emma Prest, Maria Fernanda Cruz Chaves, Brian Fitzpatrick and Sophia Baumann, with Karina Shedrofsky and Jeremie Baruch. FCIU document: 21 SARs, around 3,000 transactions and $581.9 million. Offboarding and Investigations sections: March 2020 email, references and final account closures. Internal records described by the reporters, without complete published facsimiles.

  7. S07 · ZDF frontal · Published: 2026-11-17 · Accessed: 2026-11-18 Bank references and the monitors’ warning
    Passages and scope

    Sophia Baumann, Jonas Halbe and Daniel Laufer. Reporting on references, the relationship manager’s knowledge and monitors’ concerns about other cases. Records described by journalists with access to the corpus.

  8. S08 · Der Standard · Published: 2026-11-17 · Accessed: 2026-11-18 Final account chronology and internal investigation
    Passages and scope

    Gives 9 July 2019 as the date of final account closures. Partner publication of the same collective investigation, rather than independent corroboration of the consortium.

  9. S09 · Inside Deutsche Bank · Published: not specified · Accessed: 2026-11-18 Data origins, methodology and access to the leak
    Passages and scope

    The Files and The Data sections: functions covered, editorial independence, raw data withheld from publication and access through DARC. The consortium’s account of its own methods.

  10. S10 · ZDF frontal · Published: 2026-11-17 · Accessed: 2026-11-18 Method and timetable of the collective investigation
    Passages and scope

    Wie wurden die Unterlagen ausgewertet? section: collective work since spring 2026. Distinguishes the research timetable from the publication launch on 17 November.

  11. S11 · Deutsche Bank · Published: 2026-11-17 · Accessed: 2026-11-18 Official response to the consortium’s reporting
    Passages and scope

    The bank describes historical matters and claims a transformation, including €7 billion of direct investment in controls during 2019–2025. The institution’s account of its own actions.

  12. S12 · Bundestag · German federal government · Published: 2026-10-09 · Accessed: 2026-11-18 5 October replies on BaFin and the Wyden report
    Passages and scope

    Drucksache 21/8470, questions 22–24, printed pages 14–16: assessment of US records, jurisdiction and partially classified replies. Replies dated 5 October, published on 9 October.

  13. S13 · Bundestag · German federal government · Published: 2026-09-21 · Accessed: 2026-11-18 Scope of BaFin’s inquiries concerning shareholders
    Passages and scope

    Drucksache 21/8068, page 3, answers 1–5. Ministry response transmitted on 18 September and published on 21 September. Review of significant shareholders. Parliamentary summary dated 12 October: Official summary.

  14. S14 · l0g · Published: 2026-08-09 · Accessed: 2026-11-18 L’argent d’Epstein: edition and SHA-256 digest
    Passages and scope

    French edition. Page and download checked on 18 November: HTTP 200, 198,653 bytes. Public download digest compared with the original repository binary. The current Last-Modified value is technical metadata, separate from the original edition date.

  15. S15 · GitHub · bluetouff/l0g repository · Published: 2026-08-09 · Accessed: 2026-11-18 Repository record and 9 August EPUB build
    Passages and scope

    Run events and steps checked through the GitHub API: successful attestation at 18:22:19–18:22:21 UTC, then publication to built. Original commit binary matches the current public download. Historical bundle unavailable for renewed cryptographic verification; historical Apache activation time not established. Original article · Original EPUB · Historical workflow

  16. S16 · FinCEN · Published: not specified · Accessed: 2026-11-18 SAR FAQ: transaction aggregation and continuing activity
    Passages and scope

    Question 23, Items 29 and 31: transaction totals, deposits and withdrawals, period amounts and cumulative continuing-activity amounts. General filing guidance, without validation of the Epstein-specific calculation.

Limits

  • The leak’s raw data and complete Epstein-related internal reports were not available for our direct examination. New internal findings are attributed to the consortium’s published investigations. The consortium continues its work.
  • The $581.9 million and the earlier total above $250 million have different published descriptions. The public material examined does not permit a transaction-by-transaction reconciliation or a reconstruction of net economic flows.
  • Regulatory findings, parliamentary assessments and suspicious-activity reporting have distinct legal functions. The NYDFS’s $150 million administrative penalty also covered FBME and Danske Estonia. The order gives no separate Epstein allocation. Parts of the German supervisory response remain classified. S02 S03
  • The 100-day comparison runs from 9 August to 17 November 2026. The preserved EPUB supports textual continuity. The unrecovered historical archive and attestation bundle, and undated historical Apache activation, limit the separate claims that can be made about deployment and fresh cryptographic verification.

This analysis is not investment advice.

// cite this analysis

l0g, “Deutsche Bank and Epstein: the compliance exception in the internal record”, l0g.fr, published November 18, 2026, updated November 18, 2026, https://l0g.fr/en/analysis/deutsche-bank-epstein-inside-compliance-exception/


$ cd ../analysis