Sleuth Report: Tommy Carstensen Builds a Searchable Ledger of 90,964 Amounts Found in the Epstein Files

UpdatedSeptember 3, 2026

ByEpsteinWiki

Tommy Carstensen’s Documented Amounts ledger extracts financial figures from all 37 collections in the Epstein files corpus and connects each amount to its source document. The ledger contains 90,964 cited amounts found across 25,004 documents, including 9,942 entries that resemble transaction records.

The tool is potentially invaluable for researchers. It is also easy to misunderstand. It does not prove that every amount belonged to Jeffrey Epstein, identify the payer and recipient of every transaction, or establish criminal conduct. Instead, it gives researchers a documented starting point for finding financial evidence buried inside millions of pages.

Report Snapshot

Original report: Documented Amounts: The Epstein Money Ledger

Researcher: Tommy Carstensen

Report type: Searchable financial amount extraction ledger

Evidence base: All 37 collections in the Epstein files corpus

Total cited amounts: 90,964

Entries classified as possible transaction records: 9,942

Source documents represented: 25,004

Possible transaction entries of $1 million or more: 1,034

Public table: The largest 1,000 entries in the filtered transaction subset

Lowest amount currently included in the public table: Approximately $163,800

What the Epstein Money Ledger Does

The Documented Amounts ledger searches the Epstein files corpus for amounts accompanied by currency symbols or currency identifiers. It then evaluates the language surrounding each amount to determine whether it appears near financial terminology.

Each published row includes the amount, currency, possible financial instrument, nearby date, EFTA document number, number of documents carrying the same figure, and a short section of surrounding text.

This makes the ledger useful for locating financial evidence that would otherwise remain buried inside bank statements, wire records, court exhibits, emails, tax filings, checks, loan documents, account summaries, and corporate records.

The ledger does not convert the extracted amounts into a finished investigative conclusion. Researchers must open the linked document, read the relevant pages, determine what the amount represents, and identify whether it has any meaningful connection to Epstein or his network.

The Ledger Contains 90,964 Cited Amounts

Carstensen reports extracting 90,964 amounts from 25,004 source documents.

Most entries are denominated in United States dollars. The project currently lists:

88,560 amounts in United States dollars
1,908 amounts in euros
460 amounts in British pounds
36 amounts in Swiss francs

These figures describe the currency markers detected by the extraction process. They do not represent the total value of Epstein’s fortune or the total amount transferred through his accounts.

Adding all 90,964 entries together would produce a meaningless total because the corpus contains account balances, repeated statements, proposed investments, legal claims, market valuations, news articles, reversed entries, duplicate records, and transactions unrelated to Epstein.

Only 9,942 Entries Passed the Transaction Filter

The project classifies 9,942 of the 90,964 extracted amounts as possible transaction records.

To pass this filter, an amount must appear near banking or transactional language and have a date nearby. This is designed to move actual bank records, wire instructions, account statements, checks, and payment records ahead of unrelated financial references.

However, the filter remains an automated research aid. It is not a factual determination that a transaction occurred.

A news article discussing a financial event may contain an amount, a date, and terms such as wire or transfer. A financial presentation may discuss a proposed investment that never happened. A bank statement may show an ending balance rather than money transferred during that period.

For this reason, the 9,942 entries should be described as possible transaction records selected by an automated filter, not as 9,942 proven Epstein transactions.

The Public Ledger Shows the Largest 1,000 Filtered Amounts

Carstensen publishes the largest 1,000 entries within the filtered subset. The current public table extends down to approximately $163,800.

The full collection of 90,964 extracted amounts remains unpublished while it undergoes additional review.

This decision improves usability because the largest financial records are more likely to reveal major accounts, investment structures, asset movements, suspicious activity filings, and large transfers. However, it also means the public ledger does not yet provide a complete picture of smaller payments.

That limitation is especially important in the Epstein investigation. Payments connected to recruitment, travel, tuition, rent, immigration services, cash withdrawals, staff expenses, and individual women may be far smaller than the investments and account balances appearing near the top of the ledger.

Large numbers reveal the scale of the financial infrastructure. Smaller recurring payments may reveal how the machinery operated.

The Largest Number Is Not Epstein’s Money

The first row currently lists $10.6 billion from EFTA00316714.

The surrounding text shows that the amount concerns HSBC increasing its reserves for anticipated subprime mortgage losses in 2007. It is not an Epstein account balance or a transfer involving Epstein.

This is one of the ledger’s most important lessons. A document’s presence in the Epstein files does not mean that every person, company, event, or dollar amount inside it concerns Epstein.

The corpus includes news articles, research material, forwarded messages, corporate documents, legal exhibits, and general financial information. Some documents were collected because Epstein received or stored them. Others appear as attachments or exhibits inside larger productions.

Therefore, researchers must never present a ledger entry as Epstein’s money without first reading the underlying document.

The $951 Million Entry Is Another Warning

The ledger lists $951 million from EFTA02394241.

The context shows that the amount appeared in a message sharing an article about a cyberattack involving the SWIFT banking system. The figure concerns money stolen in an international banking incident. It does not document Epstein transferring or possessing $951 million.

The extraction worked correctly in one limited sense. It found a financial amount near banking vocabulary and a date.

The entry still does not represent an Epstein transaction.

This distinction demonstrates why automated extraction is excellent for discovery but insufficient for interpretation.

The $757.2 Million Entry Requires Its Own Investigation

A $757.2 million figure appears in EFTA02822958.

The surrounding text describes an account receiving an unusually high volume of customer wires and transfers on January 3, 2007. The document states that the amount was approximately 27 times the account’s average daily incoming wire activity.

That language makes the entry potentially significant. However, the ledger does not resolve the identity of the account holder, the origin of the funds, the recipients, or the relationship to Epstein.

Those questions can only be answered by reviewing the complete document, identifying the litigation or investigation from which it came, and tracing the account references through related records.

The ledger correctly points researchers toward the evidence. It does not complete the investigation.

A $381.3 Million Entry Appears Beside Epstein’s Name

One of the most striking entries appears in EFTA01252939.

The extracted text places $381,300,136 near language describing a funds transfer involving Crédit Agricole and a line that appears to include Jeffrey Epstein’s name. The same section also contains text that may refer to a decorative project, expenses, and a much smaller disbursement.

This entry demands careful examination because the amount could represent an account identifier, a malformed account reference, a balance, a currency conversion artifact, or a genuine financial amount. Optical character recognition can merge adjacent fields and remove punctuation from scanned records.

The amount should not be reported as a $381.3 million payment to Epstein unless the original page and surrounding records establish that interpretation.

The ledger itself does not make that claim.

Account Balances Are Not Transfers

Several of the largest entries come from account statements.

EFTA01410788 includes balances of approximately $76.97 million and $73.34 million. Other records contain balances exceeding $40 million.

An account balance documents the value held in an account at a particular moment. It is not necessarily a deposit, withdrawal, or transfer.

The same balance may appear repeatedly across multiple pages or statements. A beginning balance may become an ending balance. A market value may appear beside an unrelated fee. A single amount may be listed several times within one document.

Researchers should label these figures precisely. A $73 million account balance is not the same as a $73 million wire.

Suspicious Activity Reports Require Precise Language

The ledger includes several amounts extracted from Suspicious Activity Reports.

EFTA01656415 lists approximately $45.36 million as an amount involved and approximately $73.02 million as a cumulative amount. The record identifies suspicious electronic funds transfers or wire activity during a period beginning July 15, 2019.

EFTA01656524 identifies approximately $47.3 million in suspicious activity during a period extending from 2015 through 2019.

EFTA01656452 records approximately $27.66 million in suspicious activity during a period surrounding Epstein’s July 2019 arrest.

A Suspicious Activity Report does not prove money laundering or another crime. Banks file these confidential reports when activity meets reporting criteria or raises concerns requiring notice to the federal government.

The amounts remain important because they show the scale of activity that financial institutions considered suspicious. They must still be described as amounts reported in Suspicious Activity Reports rather than proven criminal proceeds.

The Ledger Reveals Large Balances in Epstein Related Accounts

Several entries clearly come from statements bearing the names of Epstein controlled companies.

EFTA01437262 contains account records involving Zorro Management and Southern Financial. The ledger identifies entries of approximately $31.56 million and $24.56 million in the document.

EFTA01542735 contains a JPMorgan Private Bank statement for NES, LLC.

EFTA01531826 contains a business account statement for Air Ghislaine, Inc.

EFTA01561062 contains financial information for New York Strategy Group, LLC.

These entries help researchers trace which entities held accounts, which banks serviced them, and how balances changed over time.

They also support EpsteinWiki’s broader investigation into Jeffrey Epstein’s companies, trusts, and financial infrastructure.

The Records Show the Value of Entity Level Research

Epstein’s financial activity was distributed across companies, trusts, foundations, property entities, aircraft companies, and personal accounts.

The ledger contains records associated with entities including NES, LLC, JEGE, Inc., HBRK Associates, Air Ghislaine, New York Strategy Group, Zorro Management, Southern Financial, Epstein Interests, and several real estate companies.

A researcher searching only for Jeffrey Epstein’s name could miss transactions recorded under these entities.

This is why the ledger becomes more powerful when used with Carstensen’s Epstein money network and EpsteinWiki’s article on Epstein’s shell companies.

The network shows relationships between people and entities. The ledger locates individual financial figures. The original EFTA documents provide the evidence required to verify each interpretation.

Duplicate Records Are Collapsed

The Department of Justice production contains repeated and near identical records. A single bank statement or exhibit may appear in several collections or copies.

Carstensen reports collapsing 5,114 repeated rows from the public ranking.

The Docs column shows how many separate documents contain the same amount. A higher number does not mean that the money moved several times. It means the amount appears in several documents.

This distinction prevents researchers from accidentally multiplying the same balance or transaction.

Even after automated duplicate removal, researchers should examine whether similar entries represent the same transaction, repeated statements, revised filings, beginning and ending balances, or genuinely separate transfers.

The Instrument Categories Are Research Guides

The ledger classifies extracted entries according to nearby financial vocabulary.

The current classifications include:

Transfer or payment language near 40,183 entries
Account statement language near 21,830 entries
Wire or funds transfer language near 15,471 entries
Invoice language near 10,322 entries
Tax filing language near 6,087 entries
Ledger or journal language near 5,558 entries
Check or draft language near 4,518 entries
Loan or note language near 2,144 entries

These categories overlap. One row may be counted under multiple instruments.

The surrounding context can also reach a neighboring line. Therefore, the presence of wire language does not guarantee that the extracted amount is the amount of the wire. It may be a balance, market value, fee, account total, or unrelated figure appearing nearby.

How the Extraction Process Works

The project captures an amount when it includes a currency marker and appears near financial vocabulary.

Bare numbers without a currency indicator or financial context are excluded. Long strings that resemble account or card numbers are rejected. Amounts above $100 billion are treated as probable extraction errors.

The tool also requires scale markers such as M or B to be directly attached to the number. This reduces the chance that a letter appearing elsewhere in a bank record will incorrectly convert an ordinary amount into millions or billions.

Context sections are screened for personal information. Protected documents are excluded before extraction.

These safeguards reduce obvious errors. They do not eliminate errors created by poor scans, broken tables, misplaced decimal points, optical character recognition failures, or language pulled from adjacent lines.

The Ledger Refuses to Guess Who Paid Whom

One of the most responsible choices in the project is its refusal to automatically assign payer and recipient relationships.

Bank records contain complicated layouts. Names, account numbers, dates, credits, debits, and reference text may appear in separate columns. When those records are converted into plain text, their original relationships can disappear.

An automated system might confidently identify the wrong person as a payer or recipient simply because a name appears near an amount.

Carstensen therefore publishes the figure, context, date, and source without claiming to have resolved every transaction.

Researchers must return to the original document to determine the direction of the payment and the identities of the parties.

The Ledger Does Not Allege Racketeering

The report expressly states that the extracted amounts do not establish racketeering or criminal conduct.

This caution responds to a real legal issue. A large network of banks, companies, payments, and suspicious activity may look like a criminal enterprise in a visual map. However, racketeering is a specific legal claim with elements that must be proven.

In litigation brought by an Epstein survivor against Deutsche Bank, the plaintiff asserted federal civil racketeering claims. Judge Jed Rakoff dismissed those claims in May 2023. The court concluded that the complaint did not adequately allege that Deutsche Bank directed the affairs of the trafficking enterprise or consciously agreed to participate in it.

The United States Virgin Islands also asserted a territorial racketeering claim against JPMorgan. That claim was dismissed because the court found that the complaint did not adequately allege that JPMorgan conducted Epstein’s trafficking venture.

Those dismissals did not absolve the banks of every alleged failure. Other claims continued, and both banks later entered major settlements. However, the settlements were not findings that the banks had committed racketeering.

Bank Settlements and Regulatory Findings Remain Significant

Deutsche Bank agreed to a $75 million settlement with Epstein survivors. JPMorgan agreed to a $290 million settlement with a survivor class and a separate $75 million settlement with the United States Virgin Islands.

The settlements did not include admissions that the banks had participated in racketeering.

Separately, the New York State Department of Financial Services imposed a $150 million penalty on Deutsche Bank for compliance failures involving Epstein and other high risk banking relationships.

The regulator found that Deutsche Bank failed to properly monitor Epstein’s account activity despite knowing his criminal history. It identified payments to alleged co conspirators, settlements, legal expenses, Russian models, tuition, hotels, rent, women with Eastern European surnames, and more than $800,000 in cash withdrawals.

Those are regulatory findings contained in an official consent order. They are different from the dismissed racketeering allegations and should be reported separately.

Key Takeaways

The ledger contains 90,964 financial amounts extracted from 25,004 documents across all 37 collections in the Epstein files corpus.
Only 9,942 entries passed the project’s automated transaction filter.
The public page currently publishes the largest 1,000 entries, extending down to approximately $163,800.
The amounts cannot be added together to calculate Epstein’s wealth or total financial activity.
Not every amount concerns Epstein. The corpus includes news reports, forwarded articles, unrelated legal exhibits, financial presentations, and general reading material.
An account balance is not the same as a payment or wire transfer.
An amount listed in a Suspicious Activity Report is not automatically proven criminal proceeds.
Duplicate appearances do not establish that the same amount moved several times.
Automated extraction can locate evidence, but researchers must inspect the original document before identifying a payer, recipient, purpose, or connection to Epstein.
The ledger does not allege that every person or institution appearing in the documents committed misconduct.
The financial records are most useful when combined with entity research, corporate records, account histories, litigation documents, and the Epstein money network.
Smaller payments may be more important to understanding recruitment and trafficking operations than the enormous balances appearing at the top of the public table.

Fact Check Assessment

Carstensen accurately describes the ledger as an index of amounts rather than a completed record of Epstein transactions.

The project’s warnings are supported by examples visible in its own table. The $10.6 billion HSBC reserve, the $951 million cyberattack reference, and several general investment figures demonstrate that a correctly extracted amount may still have no direct connection to Epstein’s money.

The ledger’s transaction filter improves relevance but does not independently prove that each entry represents a completed transaction.

The report is also correct to distinguish financial relationships from racketeering findings. The federal and territorial racketeering claims described on the page were dismissed. Later bank settlements did not rest on findings that the institutions had committed racketeering.

The strongest use of the ledger is document discovery. Every material claim must still be verified against the linked EFTA record and, when necessary, the original page image.

Why This Ledger Matters

The Epstein files contain an overwhelming volume of financial material. Bank statements, wire instructions, checks, account summaries, investment records, trust documents, tax records, and court exhibits are scattered across millions of pages.

The ledger turns those pages into leads.

A researcher can locate an amount, identify its source document, inspect the surrounding language, and begin tracing the relevant person, bank, account, company, or trust.

That does not replace human investigation. It makes human investigation possible at a scale that would otherwise be unmanageable.

The ledger’s greatest strength is also its clearest warning. It can tell us where a number appears. It cannot tell us what the number means until someone reads the evidence.

Sources

PostedAugust 5, 2026