Epstein Data Wiki

Ugo Brachetti Peretti

Properties2
tagsperson, api, anonima-petroli-italiana, italy, oil-and-gas, jeffrey-epstein, correspondence, paris
aliasesBrachetti Peretti, Brachetti Peretti, Ugo, Ugo Brachetti, Dott. Ugo Brachetti Peretti, Dott Bracchetti, UBP

53 min read

Named on the documents as”Ugo Brachetti Peretti” and “Brachetti Peretti, Ugo” in the sender field of his own messages12; “Ugo Brachetti” in his own sign-offs34; “ugo” and “U” in others15; “UBP” as the whole sender-field display name on one sheet he signs “Ugo Brachetti”4; “Dott. Bracchetti”, with a double c, in the correspondence of a third company’s staff and on Epstein’s own schedule67; “Dott. Ugo Brachetti Peretti” on a contact card in the release8
The title the documents give him”Presidente api anonima petroli italiana”, printed in the signature block beneath his own name. Two of the sheets carrying it were read at 400 dpi910; it stands on at least two more, read from the text layer, which damages the words on both115; “Presidente” against the organisation “API Anonima Petroli Italiana” on the contact card8. No document read for this page states a shareholding, a directorship or any other office
Who he corresponds withJeffrey Epstein directly, and Epstein’s assistant Lesley Groff. His own office — a secretary who signs “Segreteria di Presidenza” at api — corresponds with the administrative assistant to Eduardo Teodorani-Fabbri of CNH Industrial N.V. about a Paris meeting12
Who Epstein passed his company toTom Pritzker, 26 June 20149; Ehud Barak, 2 July 20141314; Joshua Cooper Ramo, by an introduction e-mail addressed to both men, 15 October 20151516
Who came to Epstein about itAriane de Rothschild, whose bank was working the file under a code name by 4 March 2015 and who asked Epstein whether her man could approach the subject directly1718
What the correspondence is aboutThe sale of, or a partner for, the downstream business of the api group — a valuation, a five per cent commission Epstein proposed for himself, and a series of Paris meetings11920
Documents he wrote, cited here, run from19 June 2014, the date printed on his first message sending Epstein a presentation of his company, to 19 June 2018, the last message in the set he wrote. A few documents outside that range are cited for one of three reasons: to bound a search, to say what is left to another page, or — in the case of the 29 March 2019 chat log — to give the provenance of the contact card the title row above relies on21422
Response to the correspondence becoming publicNone located. See Response

Ugo Brachetti Peretti is an Italian oil executive who, between 2014 and 2018, wrote to Jeffrey Epstein about finding a buyer or a partner for his family’s fuel and refining business. The signature block beneath his name on the sheets reads “Presidente api anonima petroli italiana”.9 The correspondence is business correspondence: a company presentation, a valuation, meetings in Paris, and Epstein’s own proposal that he take “five percent of all money’s received” for his help.1

He is on this page because the documents record him acting, not because he is mentioned. Of the 140 documents returned by the widest search on his surname, 49 carry his name at the head of the sender field and 54 carry it in the sender field at all.23 The rest of the population is Epstein writing to him, Epstein’s office arranging his meetings, and Epstein passing his company on to other people.

Nothing on this page is a finding of wrongdoing, and no document cited here charges him with anything. What the documents show is a man trying to sell a company and a financier offering to broker it. Not everything in the set is about the company, and what is not is set out in its own section below rather than folded into the business narrative.

Identity and account numbers, telephone and fax numbers, e-mail addresses, residential addresses and dates of birth printed on the sheets cited here are not reproduced on this page.

What was searched, and what each search returned

Every figure below was measured against the FTS5 index of the full-text corpus on 11 September 2026, counting rowid first and resolving identifiers separately. Page counts are pages; document counts are distinct identifiers.

querypagesdocuments
"Brachetti"173140
"Brachetti Peretti"153122
"Ugo Brachetti"139109
"Ugo Brachetti Peretti"131103
"Brachetti Peretti Ugo"1616
"Anonima Petroli Italiana"2121
"Presidente api"1414
"gruppoapi"2823
"Gruppo API"54
"Ferdinando Brachetti"44
"Nicole Brachetti"33
"w/ugo"1313
"Bracchetti" (double c)3325
"apiX"1212

"Brachetti" at 140 documents is not the whole record, and this page says so rather than reporting it as the population. Three other searches reach documents the surname search cannot.

The double-c spelling. "Bracchetti" returns 25 documents, 16 of them outside the 140. They are not damage: they are how the staff of a third company, CNH Industrial N.V., spelled the name in their own correspondence about a Paris meeting in June and July 2014, and the spelling then propagated into Epstein’s office and onto his own schedule. All 16 were opened; all 16 concern this subject. The clearest is a reminder on Epstein’s September 2014 schedule: “Reminder: Eduardo wants you to meet with Dott. Bracchetti …said it could probably happen in Sept.”6

The code name. "apiX" returns 12 documents, 11 of them outside the 140. Ten are unrelated: bank statements, FedEx invoices, tax notices and a trial exhibit, in which the string is an artefact of the extraction. One is real, and it matters: Ariane de Rothschild’s own message to Epstein of 4 March 2015, forwarding an internal note from her bank under the subject “apiX”, which the message Epstein sent back a minute later quotes in full.17 All 12 were opened.

"gruppoapi" returns 23 documents, 8 outside the 140; seven of those eight belong to the Nicole Junkermann correspondence and are not used here, and the eighth is a June 2014 message in Epstein’s office about the same Paris meeting.24

The union of the surname population, the double-c documents and the two additional documents is 158 documents. That is the population this page was written from.

A forename alone is not a query. "Ugo" searched across the whole corpus returns 303 pages in 265 documents, of which only 112 are inside the 140; 153 documents carry the forename and have nothing established to do with this subject. Every figure on this page keyed on the forename is keyed on it within the surname population, never across the corpus.25

The surname belongs to a family. Within the 140 documents, the token printed immediately before “Brachetti” is Ugo 265 times in 109 documents, Ferdinando 8 times in 4, Nicole 3 times in 3, and Sylvianne once.26 The three Nicole documents are one thread of January 2019 and belong to Nicole Junkermann, who married into the surname in 2017; nothing from that thread is used on this page. See What this page leaves to another page.

The introduction, June and July 2014

The earliest document in the set is not from him. In mid-June 2014 an administrative assistant at CNH Industrial N.V. wrote to two secretaries, one of them at api and one of them Epstein’s, proposing dates: “Eduardo has asked me to organise a meeting in Paris next week with Dott. Bracchetti and Mr. Epstein.”27 The officer she names is Eduardo Teodorani-Fabbri, identified by name on her own signature block as Senior Vice President, Corporate Development at CNH Industrial N.V.27 On the api side the reply came from a secretary who signs “Segreteria di Presidenza”: “Mr. Brachetti Peretti ask to organize the meeting on Wednesday evening or Thursday morning early - no later than h. 10 - because he have to fly back to Rome at h.13.”12

On 19 June 2014, the same day as that exchange, he wrote to Epstein directly for the first time in this set: “Hi Jeffrey, / as agreed here is a brief presentation of my company. / thank you / all the best / ugo”, attaching a file named Presentazione_Gruppo_June_2014_v1.pdf.21 He sent it again on 26 June under the subject “api group presentation”.11

What Epstein did with it, twice, within a week, is the first real finding on this page.

On the evening of 26 June 2014 he forwarded it to Tom Pritzker with a covering line of his own: “itailin oil , however i think that rothschild is a better deal. for you. I would put up as much as you want. stock down, great franchise. needs operator. , great platform for you. ( israel. france swiss. )”.9

On 2 July 2014 he forwarded it to Ehud Barak with two words: “for victor?”13 Barak answered the same morning: “Jeff / I’m in NY on 16-20 July. / Viktor will probably be in NY as well at least part of this period. I’ll try to reset our tri lateral meeting. / Re this , let me think. Probably VV. Probably a PE fund? An oil Bigger player? / Let’s talk about it later today or tomorrow. / EB”. Epstein replied “i can be there the 18th only”, and Barak: “Ok. I’ll try to seduce him into a meeting on this date. / Best / EB”.14 No document read for this page identifies “victor”, “Viktor” or “VV”, and this page does not supply a name for them. No document read for this page records any meeting on 18 July 2014 or any answer from Pritzker.

Epstein’s own line to Brachetti Peretti at the end of that week was: “ramadan is begiining but there might be some interest I will be in paris week of july 20th. hope to see you there.”28 The Paris meeting did not happen. On 20 July Brachetti Peretti wrote “hi, / could we meet in Par=s tuesday night? / Ugo Brachetti”, and Epstein answered: “sorry , but im working on a very large transaction , that will need me in wash dc this week, will let you know when i can come or im happy to see you in the states”.329

On the = characters. The sheet at EFTA01802655 prints “Par=s” and, further down, “Ok i will organize to come =n Paris in that week”. The = is on the image, not introduced by the extraction: it is quoted-printable transmission encoding surviving onto the printed face of that production copy, and it is not repaired here.3 A second copy of the same message at EFTA00642505 prints the same words undamaged, “could we meet in Paris tuesday night?”, so the damage is one copy’s, not the message’s.30

”your arab friend”, October 2014

On 7 October 2014 he wrote again, and this is the sentence that states what he wanted:

Hi Jeffrey,hope everything is well with you,i was wondering if there is still the possibility to discuss about my company getting in touch with your arab friend to talk about potential partnership.I m eventually available when you want

2

Epstein answered four minutes later: “of course. i will be in paris on the 22”, and Brachetti Peretti: “Ok i can join you there / We keep in touch”.31 On 20 October he asked “Hi Jeffrey / you r going to be in Paris in the 22?”; Epstein’s whole reply was “27yth”; his whole reply to that was “ok”.32 No document read for this page names the person Epstein’s correspondent called “your arab friend”, and this page does not supply one. The chain carrying the phrase appears in 15 documents of the population, and in none of them does a name stand beside it.33

Paris, February and March 2015

The meeting that did happen took eight days to arrange, and the arranging is in the release in full. On 24 February 2015 Epstein’s assistant Lesley Groff opened it: “Hello Ugo. Jeffrey will be in Paris starting on Feb. 28th for the week…might you have some time to come by and see him at his Paris apartment? Perhaps Monday March 2nd? Or feel free to suggest a day/time…”5 He proposed Wednesday 4 March; the time then moved from 12:30 to 1:00, then to 3:00, then to 1:30. “Yes, Lunch would be great. Does 12:30pm work for you at Jeffrey’s home?” / “1pm would be better” / “Sure. 1:00 will be great.” / “Jeffrey can’t meet with you at 1pm on Wed. March 4th… asking if you could meet him at 3pm instead of 1pm” / “i do not understand the timing and the day” / “ok,of he can for dinner on the 3 night or the 4th at 3 pm(for 1 hour)i have a flight at 6pm same day”.534

Epstein’s own daily schedules track the same movement. Copies in the release print, for Wednesday 4 March 2015 in Paris, “1:00pm LUNCH Appt w/Ugo Brachetti Peretti”, “3:00pm Appt w/Ugo Brachetti Peretti” and “1:30pm Appt w/Ugo Brachetti Peretti” on different versions of the same day.35 The place he was told to go is Epstein’s own property: “Please meet Jeffrey at 3pm on Wed. March 4th at his Paris Apartment. / 22 Avenue Foch, Apt 2DD”.36 It was settled on 3 March, when Groff wrote “Ugo, Jeffrey says you may go see him tomorrow, Wed. March 4, at 1:30pm. OK?” and he answered “ok i will be there”; she then wrote to Epstein, “Ugo will come see you tomorrow, Wed. March 4 at 1:30.”3637

The meeting took place, and the release records both sides saying so. At 14:56 the same afternoon he wrote:

Dear Jeffrey, thank you for todays meeting,it was great to see you after few years. i could only give you a brief presentation of my company,but there is still lot to look at. Any other questions let me know. hope to see you soon best Ugo

Groff answered at 20:06: “I have sent off to Jeffrey for you! Thanks and very happy your meeting worked out. Lesley”.38 The words “after few years” are his own, and they place a first acquaintance before this correspondence begins. No document read for this page dates it, records how the two men met, or records any contact between them before June 2014.

4 and 5 March 2015: the valuation

What follows the meeting is the strongest sequence in the set, and it runs over about twenty-eight hours.

4 March, 10:28. Epstein: “questions re what is your brothers role and view in this search. ??”39 11:02. Brachetti Peretti: “they totally agree this potential solution. i will ex pane you in details when we see each other / im in town if you want to anticipate the meeting its fine for me / Ugo”.39

4 March, 14:29. Ariane de Rothschild forwarded Epstein an internal note written that morning by a banker at Edmond de Rothschild, under the subject “apiX”, which opens “Sorry for the appalling code name (although it is a reference to our first African PPP!)” and goes on: “In addition to a large fund and/ or industrial sponsor who would have an interest in the development strategy of the company, one route could also be to pool a group a Italian institutions/ large deep pocket families… Which apiX offers as it provides both Italian integrated energy assets and Italian/ international renewable development strategy.” Her own covering line was: “Fyi / Do you think my guy could go meet Ugo ( or his cfo) directly?”17

14:30, one minute later. Epstein answered: “of course. ,”, then a value covered by a bar, then ”— Ugo Brachetti Peretti”.18 The extraction of that sheet closes the bar and prints the two halves as consecutive lines; at 400 dpi the bar is on the image, and what it covers is not supplied here.

4 March, 20:07. Epstein to Brachetti Peretti: “ariane will follow up ,”, then a value covered by a bar, then “will follow up.. I was unclear how you came to a value of 1 billion for a co that curently has an ebitda of 78? . or did i miss something”.20

5 March, 09:04. Brachetti Peretti answered the question with a memorandum, forwarded under the subject “downstream sector valuation”:19

Hi Jeffrey, here is the answer to your question,and some explanation on the evaluation of the company. Note that 2014 represent the first full year after the turnaround of the group in 2013; the downstream sector achieved an ebitda of 80m which is double the one of the previous year. we expect a natural increase of the marginality of the business once all actions currently ongoing will be finalized up to 140m.

Valuations based on sum of the parts/comparable market transactions

in a sector where the volatility of the results is important we normally use as a reference point for valuations an approach related to sum of the parts based on market references or previous comparable transactions. in evaluating api group downstream activity we highlight the following items: refinery 200-300m logistic/storage 200-250m retail network (owned) 600-700m retail network (partner) 150-200m wholesale unbranded 50m wholesale special products 250-300m

total Enterprise Value 1450-1800m

5 March, 12:54. Epstein’s reply is the document the rest of this page turns on:

Ugo , I m not sure of much but am certain that the valuation method is very very high . If you want my help , Indians Chinese Arabs . The deal is five percent of all money’s received . Debt equity etc . I don’t need a contract only a paragraph , if not I won’t be insulted and we can talk women instead .

1

14:11. Brachetti Peretti: “Jeffrey, / on the base of your proposal,how should i behave with Ariane Rothschild that just wrote to me to discuss the possibility to assist us on the possible identification of a partner? / thanks / ugo”.1 14:22. Epstein: “I’ll talk to her sat”. 14:24. Brachetti Peretti: “ok then i will wait to contact her”.10

What that thread establishes and what it does not. It establishes, in Epstein’s own typed words to a principal, that he offered to find a buyer for an operating European company and named his own fee: five per cent of all money received, on debt or equity, against “only a paragraph” rather than a contract. It establishes that his correspondent treated the offer as live, in the words “on the base of your proposal”, and deferred to him on how to handle an approach from a bank. It establishes that a Rothschild bank was working the same file under a code name on the same day, and that Epstein was in both conversations at once.

It does not establish that any agreement was made. No document read for this page records a signed engagement, a paragraph of the kind Epstein asked for, a payment, or any transaction resulting from any of it; the searches that would refute that are set out below.33 One introduction Epstein did make is recorded, in October 2015, and is described in the next section. No document read for this page records Brachetti Peretti accepting the five per cent, refusing it, or answering it at all: his next message changes the subject to Ariane de Rothschild.

After March 2015

The correspondence thins sharply. On 15 October 2015 Epstein made an introduction by e-mail in one line, addressed to two people at once: “ugo josh, josh ugo, ugo josh is in touch with chinese buyers intersestd in gas stattions”. The To: field of that message carries two names, “Joshua Cooper Ramo” and “Ugo Brachetti Peretti”, so the “josh” of the introduction is named on the sheet and not inferred.15 Eleven minutes later Joshua Cooper Ramo answered both of them: “Thanks Jeffrey. / Ugo, nice to meet you. How best to connect? / Joshua Ramo”.16 Brachetti Peretti’s own reply to Epstein, sent in the same minute, was “hi / let me know / ugo”.40 No document read for this page records what, if anything, followed the introduction.

On “gas stattions”. The image of EFTA00639410 prints “gas stattions”, with two ts. The extraction of that sheet prints “gas stallions”. The word on the sheet is the misspelling, not the animal; the extraction manufactured the other reading, and that is why it was checked at 700 dpi before being quoted.40

On 23 January 2018 the whole exchange was: Epstein, “I’m in Paris - you ?”; the reply, “rome”.41 The last message in the set that he wrote is of 19 June 2018, and its sender-field display name is the three letters “UBP”:

hi Jeffrey hope everything is well. Eduardo told me about a dinner that you are having in Paris with Italian rappresentative on Sunday. i will be in Paris the we. hope to see you best Ugo Brachetti

4

No document read for this page records a reply to it.

The latest-dated document in the population is neither his nor about the company. On 29 March 2019, nine months after that message and nearly five years after that same man’s office had arranged the June 2014 introduction, Eduardo Teodorani-Fabbri sent Epstein an attachment, timed on the sheet at 10:23:17 am and named there Dott. Ugo Brachetti Peretti.vcf. That sheet is one page of a device extraction of a private conversation between those two men, running from 18 December 2018 to 30 March 2019. The subject of this page is named on it once, in that filename, and nothing else in that conversation is used here. The contact card produced at the next page stamp in the same run is the card the infobox above relies on for the printed title.22

What is in the correspondence besides the company

A page that reported only the business would be incomplete in two places, and both are set out here.

The closing clause of the fee message. Epstein’s message of 5 March 2015 ends “if not I won’t be insulted and we can talk women instead”. It is quoted in full in the section above rather than trimmed, because it stands in the same message as the five per cent and cutting it would misreport what the sheet says.1

23 June 2015. On the evening of that day Epstein wrote to him: “do you know / Chiara Baschetti.. milan cute”, the name carrying a link to a modelling-gallery page about her. The reply was “No,how can i help?” Epstein: “she was shooting a commercial in front of my ny house. very cute”. The reply: “She is,i will find out about her”.42 An hour later Epstein wrote to a different correspondent, whose display name and address are both barred: “ask that girl chiara brachetti if she wants to have dinner with us and woody allen in paris”.43 The surname in that message is not hers; it is the surname of the man he had been writing to an hour earlier.

That message was answered. At 22:23 a correspondent whose sender field is wholly barred replied “Will do now. When approximately?”, and at 22:45 Epstein answered “friday july 3”.44 Whether the invitation was passed on, and what became of it, is not recorded by any document read for this page.

And the question was not put to him alone. Ten weeks earlier, on 10 April 2015, Epstein had sent the same two lines to Eduardo Teodorani-Fabbri of CNH Industrial, the officer whose office had arranged the June 2014 introduction; the answer there was “Not yet”.45 On 19 March 2016 Epstein wrote to her directly, in a message of one line: “i wondered if you have become more brave”.46

"Baschetti" returns 15 documents across the whole corpus. All 15 were opened. None of them carries a message from her, and no document read for this page records a meeting, a dinner or a reply from her.46 Nothing here is a statement about her conduct; what these sheets record is what Epstein asked of two of his correspondents, and what they answered. The absence is an absence of anything from her, and it is not a finding that nothing was done: whether any of what Epstein asked about her went further than the asking is a question these fifteen documents raise and do not answer.

What this page leaves to another page

Nicole Junkermann is a separate subject with her own page, and nothing from her correspondence is used here. Specifically left there: the January 2019 “Parcel for you” thread, which is the only place Nicole Brachetti Peretti appears in this release; the June 2017 exchange between her and Epstein that the gruppoapi search reaches, in which the gruppoapi string belongs to a third person’s address that she passed on inside the thread and not to an address of hers; and everything about her businesses.

Two matters sit on the boundary and are recorded here because the documents that carry them are in this subject’s own population.

On 17 February 2017 Epstein wrote: “I met fernando last week, he came with nicole to palm beach”. The reply the next morning was: “Hi Jeffrey, / hope everything is well, i knew my brother was around,let’s keep in touch / ugo”.47 The documented fact is that Epstein named a “fernando” and that Brachetti Peretti answered by calling that person his brother. Four other documents in the surname population print “Ferdinando Brachetti Peretti” in full, and the outside record gives Ferdinando Brachetti Peretti as his brother and as Nicole Junkermann’s husband; no sheet read for this page states the relationship between the two names, and this page does not close the identification from the corpus alone.48

On 6 June 2017 a message in the same population is from Nicole Junkermann to Lesley Groff, introducing Ferdinando Brachetti Peretti; it is about a painting, and it belongs to her page.49

Response

No response from Ugo Brachetti Peretti, from api / Italiana Petroli, or from anyone on his behalf to this correspondence becoming public has been located. Searches of the outside record for a statement, denial or comment by him or by the company returned Italian press coverage of the correspondence in February and April 2026 and no response within it; the article of 14 April 2026 carries no statement from him, his representatives or the company in the text that can be read without a subscription.50 The register rule of this wiki is that a response missing beside the allegation it answers is a defect, so the absence is recorded rather than passed over. This page does not assert that no such response exists; only that none was found by the searches named here.

What the outside record adds, and where it stops

The outside record, and not the documents, gives the company’s later history: Anonima Petroli Italiana S.p.A. became Italiana Petroli S.p.A. in 2019, and the Brachetti Peretti family agreed in September 2025 to sell it to the Azerbaijani state oil company SOCAR, a sale reported as completed in 2026.50 It also gives Ferdinando Brachetti Peretti as his brother. None of that is on a sheet cited here and none of it is used as evidence for anything above. It is recorded because it supplied the search terms "Italiana Petroli", which returns 0 documents, and "SOCAR", which returns 2 documents, neither concerning this subject. That is the only use this page makes of it.

What the documents do not establish

  • They do not establish his corporate role beyond the printed title. “Presidente api anonima petroli italiana” is what the signature block says, and that is all this page claims. No shareholding, no board seat, no ownership percentage and no other office is stated by any document read for this page.
  • They do not establish that Epstein ever acted on the five per cent, was ever engaged, or was ever paid. A disconfirming document would carry an engagement letter, a mandate, a retainer, an invoice, a commission or a fee. Across all 158 documents of the population, matched as whole words, commission, fee, fees, engagement letter, mandate and retainer appear in none. The footnote states the matching rule it used, and records separately what a substring search returns instead: 28 documents for fee, every one of them those three letters standing inside a longer string and none of them a fee. The test can be re-run as it was run. percent, per cent and 5% appear in six, four of which are copies of the 5 March thread quoted above and the other two contact lists in which the match falls inside a telephone string; agreement and contract appear in six, all of them the 5 March thread and the two apiX messages; invoice, wire, payment and paid appear in one, a contact list; signed and signature appear in two, neither of them a signed instrument.33
  • They do not record a meeting between the two men after 4 March 2015. Later messages propose meetings; no document read for this page records one taking place. A disconfirming document would be a schedule entry naming him after that date; in all 13 documents that carry a schedule entry naming him, that entry is for Wednesday 4 March 2015.35
  • They do not record a reply to his last message. A disconfirming document would be a message from Epstein to him or to “UBP” after 19 June 2018. Four documents in the population carry a date later than 19 June 2018 and none of them is such a message. Three are the messages of a single thread of 31 January 2019 that belongs to another subject. The fourth is a device extraction of a chat log between Epstein and Eduardo Teodorani-Fabbri running from 18 December 2018 to 30 March 2019, which is the latest date carried by any document in the population; the subject of this page is named on it once, in the filename of a contact card sent as an attachment, and it is described in the section above.33
  • They place him at no property of Epstein’s except the Paris apartment, and record no travel with him. The only address in these sheets at which he was asked to meet Epstein is the Paris apartment named in the meeting arrangements, and no document cited here records the two men travelling together or records him at any other Epstein property. Being in an address book is not a fact about conduct: his surname appears as an entry in Epstein’s contact lists, beside hundreds of others.51
  • They do not make the correspondence wholly commercial. One exchange in the set is not about the company at all, and it is set out in What is in the correspondence besides the company rather than left out.

Method

Populations were measured against the FTS5 index of the full-text corpus on 11 September 2026 and are reproducible from the queries printed in the table above. The 140 documents of the "Brachetti" population carry 137 EFTA identifiers, 2 HOUSE_OVERSIGHT identifiers and 1 DOJ-OGR identifier; the scope was taken from the identifier prefix and not from the dataset column.52

The sender-field figures were produced by the desk’s classifier over the whole 140, not over a sample, and the strict figure, meaning the name at the head of the sender field, is 49, against 54 for the name anywhere in that field. The five-document gap is entirely forwarded chains in which Epstein’s own header stands above his correspondent’s on the same sheet.23 Both figures are the classifier’s, and reading the population shows errors in both directions: at least two of the 54 are sheets whose sender is someone else, and at least three documents he plainly wrote fall outside it because the sender field carries “UBP”, or the extraction has damaged the name.23

Where a finding on this page turns on what a sheet prints, the sheet was rendered and read at 400 dpi or higher, and the footnote says so. Where it does not, the footnote says the reading is from the text layer. The text layer of this set is demonstrably unreliable on individual words: it prints “stallions” for “stattions”, “alien” for “allen”, “to” for “te”, and “bank” for “barak”; on one copy of the October 2014 chain it drops a letter from “friend”.33 In two places it closes a redaction bar and prints the two halves of a line as though nothing had been withheld, and on a third sheet it prints the From: label with nothing after it and hands the To: value to the sender slot, which is how a document written to this subject can be made to look, in the text layer alone, as though he had written it.23 So every quotation on this page that carries evidential weight was checked at the image.

Footnotes

Footnotes

  1. https://epstein-data.com/EFTA00862963, printed p.1 of 2. From: "Brachetti Peretti, Ugo" over a barred address; To: Epstein; Subject: “Re: downstream sector valuation”; Date: “Thu, 05 Mar 2015 14:11:20 +0000”. Read at 400 dpi. The sheet carries three messages: his 14:11 question at the top, Epstein’s 13:54 CET reply quoted beneath it, and his own memorandum quoted beneath that. Printed p.2 of the same document carries the signature block, where the extraction prints the surname as “Pcrctti”. Epstein’s message also stands alone at https://epstein-data.com/EFTA00862961, timed 12:54 UTC. ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7

  2. https://epstein-data.com/EFTA00997584, printed p.1 of 1. From: “Ugo Brachetti Peretti” over a barred address; Date: “Tue, 07 Oct 2014 12:49:55 +0000”. Read at 400 dpi. The line breaks in the block quotation are the sheet’s own. ↩ ↩2

  3. https://epstein-data.com/EFTA01802655, printed p.1 of 2. Sender field “Ugo Brachetti Peretti”; Sent: “Sunday, July 20, 2014 5:24 PM”; To: “EPSTEIN JEFFREY”. Read at 400 dpi, which is how the = characters were established to be on the printed face rather than introduced by the extraction. The forwarded message beneath is dated on the sheet “11 luglio 2014 08:0=:31 CEST”. ↩ ↩2 ↩3

  4. https://epstein-data.com/EFTA00946658, printed p.1 of 1. From: “UBP” over a barred address; To: “EPSTEIN JEFFREY”; Subject: “paris”; Date: “Tue, 19 Jun 2018 20:01:56 +0000”. Read at 400 dpi. A second copy is at https://epstein-data.com/EFTA02562545, whose extraction renders the subject as “pans”. Neither is counted in the 54, because the sender-field display name is the three letters and not the name. ↩ ↩2 ↩3 ↩4

  5. https://epstein-data.com/EFTA00352480, of 25 February 2015 at 17:29:51, printed pp.1–3 of 3; from “Ugo Brachetti Peretti” to Lesley Groff, carrying the whole arrangement chain from 24 February quoted beneath, with Groff’s name printed at each level. The opening message of the chain, quoted in the text above, stands on printed p.3, as does his signature block, whose title line this extraction prints as “Presidente a ’ anonima petroli italiana”. Text layer. ↩ ↩2 ↩3 ↩4

  6. https://epstein-data.com/EFTA00284982, printed p.1, and https://epstein-data.com/EFTA00363963, printed p.1, both copies of Epstein’s September 2014 schedule and both carrying the quoted reminder. Text layer. Neither is in the "Brachetti" population; both were reached only by the double-c spelling. "Bracchetti" returns 33 pages in 25 documents, 16 of them outside the 140; all 25 were opened and all 25 concern this subject. The 16 outside are: https://epstein-data.com/EFTA00284982, https://epstein-data.com/EFTA00363963, https://epstein-data.com/EFTA00366619, https://epstein-data.com/EFTA00367461, https://epstein-data.com/EFTA00367586, https://epstein-data.com/EFTA00502863, https://epstein-data.com/EFTA01750497, https://epstein-data.com/EFTA02101696, https://epstein-data.com/EFTA02101973, https://epstein-data.com/EFTA02102338, https://epstein-data.com/EFTA02102410, https://epstein-data.com/EFTA02102731, https://epstein-data.com/EFTA02103767, https://epstein-data.com/EFTA02584824, https://epstein-data.com/EFTA02585904 and https://epstein-data.com/EFTA02586518. ↩ ↩2

  7. https://epstein-data.com/EFTA00366508, of 17 July 2014, printed pp.1–2; subject “Fwd: Reminder: Eduardo & Dott Brachetti in Paris”, and https://epstein-data.com/EFTA01751417, the same message in another production copy. Text layer. ↩

  8. https://epstein-data.com/EFTA01617536, printed p.1 of 2, a contact card rendered as a field list: “Formatted Name / Dott. Ugo Brachetti Peretti”, “Organization / API Anonima Petroli Italiana”, “Presidente”. Text layer. The card’s address, telephone and e-mail fields are not reproduced. Its two printed pages carry the page stamps EFTA01617536 and EFTA01617537, immediately after the five printed pages of the chat log at https://epstein-data.com/EFTA01617531, on which an attachment of that filename is sent; see the next footnote. ↩ ↩2

  9. https://epstein-data.com/EFTA00992678, printed p.1 of 1. From: Epstein, To: “Tom Pritzker” over a barred address, Subject: “Fwd: Lettera presentazione gruppo api.”, Date: “Thu, 26 Jun 2014 20:51:43 +0000”. Read at 400 dpi, where the covering line, the forwarded message and the signature line “Presidente api anonima petroli italiana” all print clearly and the three lines beneath that title are covered by bars. ↩ ↩2 ↩3 ↩4

  10. https://epstein-data.com/EFTA00862967, printed p.1 of 2. From: "Brachetti Peretti, Ugo"; Date: “Thu, 05 Mar 2015 14:24:53 +0000”. Read at 400 dpi. This is not a second copy of EFTA00862963: it is the next message in the same thread, thirteen minutes later, carrying Epstein’s 14:22 “I’ll talk to her sat” and the whole of the earlier chain quoted beneath it. Epstein’s 14:22 message also stands alone at https://epstein-data.com/EFTA00862965. Printed p.2 of this document, also read at 400 dpi, is the copy on which the signature block prints “Ugo Brachetti Peretti” in bold and “Presidente api anonima petroli italiana” beneath it, twice and without OCR damage. The damaged copy at https://epstein-data.com/EFTA02508399 does not carry the title at all. ↩ ↩2

  11. https://epstein-data.com/EFTA00992679, of 26 June 2014 at 20:52:03, printed p.1 of 1; subject “api group presentation”. Text layer, and the title line is damaged in it: the extraction of that sheet prints the signature block as “Presidente a i anonima etroli italiana”. The undamaged form is read at the image on EFTA00992678 and on printed p.2 of EFTA00862967. ↩ ↩2

  12. https://epstein-data.com/EFTA00367594, printed p.1, whose signature block reads “api - anonima petroli italiana / Segreteria di Presidenza”. Text layer. Printed p.2 of the same document carries the CNH Industrial block naming the officer, and is cited at the previous footnote, where it is read at the image. The same passage is quoted back inside https://epstein-data.com/EFTA00367437 and https://epstein-data.com/EFTA02103088. ↩ ↩2

  13. https://epstein-data.com/EFTA00992022, printed p.1 of 1. From: Epstein, To: “ehud barak”, Subject: “Fwd: api group presentation”, Date: “Wed, 02 Jul 2014 10:06:09 +0000”; the whole covering line is “for victor?”. Text layer; the same two words are confirmed at 400 dpi on the reply sheet at the next footnote. ↩ ↩2

  14. https://epstein-data.com/EFTA01921385, printed p.1 of 2. To: Epstein, From: “ehud barak”, Sent: “Wed 7/2/2014 2:07:18 PM”, Subject: “Re: api group presentation”. Read at 400 dpi. The sheet carries the whole chain: Barak’s 06:32 message, Epstein’s 13:06 “for victor?” over the forwarded presentation, Epstein’s 13:49 “i can be there the 18th only”, and Barak’s 14:07 reply. The extraction of this sheet prints “ehud bank” for the second occurrence of the name and “Fm in NY” for “I’m in NY”; both are extraction faults and the image reads correctly. ↩ ↩2

  15. https://epstein-data.com/EFTA00843499, printed p.1, From: Epstein, To: “Joshua Cooper Ramo” over a barred address and “Ugo Brachetti Peretti”, Date: “Thu, 15 Oct 2015 15:28:58 +0000”. Text layer. This document was not found by the search that built the page; it was found by running the falsification test on an earlier draft’s claim that the surname of “josh” was not established. The extraction of this copy prints “gas stallions”; the extraction of the copy at the next footnote prints “gas stattions”, and the image settles it as “stattions”. ↩ ↩2

  16. https://epstein-data.com/EFTA00843500, printed p.1, From: “Joshua Cooper Ramo” over a barred address, To: Epstein and “Ugo Brachetti Peretti” over a barred address, Date: “Thu, 15 Oct 2015 15:39:21 +0000”. Read at 400 dpi. A second copy is at https://epstein-data.com/EFTA01732146. ↩ ↩2

  17. https://epstein-data.com/EFTA00863002, printed p.1, From: “A. de Rothschild”, To: Epstein, Subject: “Fwd: apiX”, Date: “Wed, 04 Mar 2015 14:29:31 +0000”. Text layer; the whole of its content is reproduced and read at 400 dpi inside the reply at the next footnote, where the forwarded header, the date “4 mars 2015 10:45:34 UTC+1” and the Edmond de Rothschild (UK) Limited confidentiality notice all print clearly. This document is not in the "Brachetti" population — it does not contain the surname — and was reached only through the code name. ↩ ↩2 ↩3

  18. https://epstein-data.com/EFTA00863004, printed p.1, From: Epstein, To: “A. de Rothschild” over a barred address, Subject: “Re: apiX”, Date: “Wed, 04 Mar 2015 14:30:58 +0000”. Read at 400 dpi. The first line reads “of course. ,” then a bar about twenty characters wide, then ”— Ugo Brachetti Peretti”, all on one printed line. The extraction of this sheet drops the bar and prints the two halves as two lines. ↩ ↩2

  19. https://epstein-data.com/EFTA02508399, printed p.1 of 1, sender field “Ugo Brachetti Peretti”, Sent: “Thursday, March 5, 2015 9:04 AM”, To: “EPSTEIN JEFFREY”, subject “Fwd: downstream sector valuation”. Text layer, and heavily damaged by quoted-printable on that copy; the text of the memorandum block-quoted above is taken from the clean printings at EFTA00862963 and EFTA00862967, both read at 400 dpi, and not from this copy. ↩ ↩2

  20. https://epstein-data.com/EFTA00863048, printed p.1 of 1, From: Epstein, To: “Ugo Brachetti Peretti” over a barred address, Date: “Wed, 04 Mar 2015 20:07:45 +0000”. Read at 400 dpi. A bar about seven characters wide stands between “ariane will follow up ,” and “will follow up..”; the extraction closes it and prints the line without it. What the bar covers is not supplied here. ↩ ↩2

  21. https://epstein-data.com/EFTA02709964, printed p.1 of 1. Sender field “Ugo Brachetti Peretti”; Sent: “Thursday, June 19, 2014 9:48 AM”; To: Epstein; subject “Fwd: Lettera presentazione gruppo api.”. Text layer, and damaged in it on both of the strings the sentence above uses: this extraction prints the attachment as Presentazione_GruppoJune 2014 vl.pdf, with spaces where the sheet has underscores and an l where the sheet has a 1, and prints a=brief where the sheet has “a brief”. Neither string is taken from it. The same message is quoted inside https://epstein-data.com/EFTA00992678, printed p.1, read at 400 dpi — where the forwarded header prints “Date: 2014-06-19 5:48 GMT-04:00”, the Attachments: line prints Presentazione_Gruppo_June_2014_v1.pdf with its underscores and its v1, and the forwarded body prints “as agreed here is a brief presentation of my company”. The attachment name and the quoted message in the text above are read from that image. ↩ ↩2

  22. https://epstein-data.com/EFTA01617531, printed p.4, printed pp.1–5 of 5, whose five printed pages carry the page stamps EFTA01617531 to EFTA01617535. It is a device extraction of an iMessage conversation, with date dividers running from 2018/12/18 to 2019/03/30, between Epstein and a correspondent whose name prints on printed p.2 as “TEODORANI Eduardo”. Printed p.4 read at 400 dpi, where the attachment entry prints five lines in one bubble: “Sent an attachment”, then “Filename: Dott. Ugo Brachetti Peretti.vcf”, then “Path: Dott. Ugo Brachetti Peretti.vcf”, then a size of 882 bytes and a type of text/vcard, timed beneath at “10:23:17 am”. The extraction of that page damages the filename line to Dolt. Ugo 8rachetti Peretti.vcf and prints the path line clean, so the name is taken from the image. Two further contact cards, for other people, are sent from the same side in the minute that follows; the entry is not singled out on the sheet. Printed p.3 was also read at 400 dpi, to establish that the date divider standing above that entry is “2019/03/29”. The surname appears on no other page of the document, and on that page only in the filename and path of that one attachment. Nothing else in that conversation is used here. The contact card at https://epstein-data.com/EFTA01617536 is the next document in the production run and carries the same name in its formatted-name field. ↩ ↩2

  23. Measured with scripts/act_rate.py over the whole "Brachetti" population of 140 documents, not a sample; the tool’s self-test was run first and passed. Loose figure — the key anywhere in the sender field: 54 of 140 (39%). Addressed only, 26 (19%); header-less first sheet with the name at its head, 2; header-less press or filing, 0; neither, 58. Strict figure — the key at the head of the sender field: 49 of 140 (35%), re-derived by splitting each multi-header sender field on its joins and testing only the first. The five documents in the gap are https://epstein-data.com/EFTA00352806, https://epstein-data.com/EFTA00863659, https://epstein-data.com/EFTA00992022, https://epstein-data.com/EFTA00992678 and https://epstein-data.com/EFTA01921385 — in each, the sheet carries two stacked headers and Epstein’s is the upper one. Two further documents returned __UNDETERMINED__ (https://epstein-data.com/EFTA00499782, https://epstein-data.com/EFTA02508287) and nine had no sender label at all. Errors found by reading rather than by the tool: https://epstein-data.com/EFTA00352069 is counted as his although its From: is wholly barred and the classifier has taken the first body line (read at 400 dpi); https://epstein-data.com/EFTA01921385 is counted as his although its From: reads “ehud barak” (read at 400 dpi). In the other direction, https://epstein-data.com/EFTA00946658 and https://epstein-data.com/EFTA02562545 carry the sender-field name “UBP” and are signed “Ugo Brachetti”, and the extraction of https://epstein-data.com/EFTA02082700 prints “go” where the sheet itself prints From: “Ugo Brachetti Peretti” clean and undamaged (read at 400 dpi); none of the three is counted. The net of the three known false negatives against the two known false positives is +1. A fourth identifier, https://epstein-data.com/EFTA02084576, was carried as a false negative by an earlier draft of this page and is not one. Read at 400 dpi, its From: field is a bar and “Ugo Brachetti Peretti” is the value of its To: field; the extraction prints the From: label with nothing after it and then hands the To: value to the sender slot, which is where “U o Brachetti Peretti” under From: in the text layer comes from. The body of that sheet is an assistant writing to him — “Ugo, Jeffrey is asking if you could come at 3pm instead on the 4th. Ok for you?” — so it is not a document he wrote, and its exclusion from the 54 is correct rather than an error. ↩ ↩2 ↩3 ↩4

  24. "gruppoapi" returns 28 pages in 23 documents, 8 outside the 140; all 23 were opened. Seven of the eight are messages between Epstein and Nicole Junkermann in June 2017 and are left to her page. The eighth is https://epstein-data.com/EFTA01798553, printed p.1 of 2, a message in Epstein’s office of 23 June 2014 forwarding the CNH Industrial proposal, whose extraction is damaged by quoted-printable throughout: it prints the sentence quoted in the text above as “Eduardo =as asked me to organise a meeting in Paris next week with Dott. =racchetti and Mr. Epstein.” Text layer. E-mail addresses printed in that extraction are not reproduced. ↩

  25. "Ugo" searched across the whole corpus returns 303 pages in 265 documents. Intersecting that set with the 140: 112 inside, 153 outside. The outside set was not read and nothing is asserted about it beyond its size; it is reported because a forename this common cannot carry a population. ↩

  26. Counted over the 173 pages of the "Brachetti" population by matching the token printed immediately before the surname, case-insensitively. Ugo 265 occurrences in 109 documents; Ferdinando 8 in 4; Nicole 3 in 3; Sylvianne 1 in 1. Remaining matches are OCR artefacts and honorifics (Dott, Mr., Don, AM, Appt). The desk’s brief recorded the split as 252/3/8; the Ferdinando and Nicole figures re-derive exactly and the Ugo figure re-derives at 265, a difference attributable to the separator class used between the two tokens. The Nicole documents are https://epstein-data.com/EFTA02628055, https://epstein-data.com/EFTA02628068 and https://epstein-data.com/EFTA02628104. ↩

  27. https://epstein-data.com/EFTA00367594, of 19 June 2014 at 12:36:19, printed pp.1–2; and its companion at https://epstein-data.com/EFTA00502863, printed p.1 of 1. The quoted sentence is on both sheets; text layer. Printed p.2 of EFTA00367594 was read at 500 dpi, where the signature block prints “Administrative Assistant to Eduardo Teodorani-Fabbri” above “Senior Vice President, Corporate Development” above a “CNH Industrial” line. The companion sheet carries the same block, where the company form reads “CNH Industrial N.V.”. E-mail, telephone and fax values printed in these blocks are not reproduced. ↩ ↩2

  28. https://epstein-data.com/EFTA00992248, of 28 June 2014 at 00:18:58, printed p.1; Epstein to Brachetti Peretti. Text layer; the same sentence is quoted back in https://epstein-data.com/EFTA00992257, his reply of 04:34 the same morning — “Let me know i will come to see you in Paris / Thank you / Ugo”. ↩

  29. https://epstein-data.com/EFTA00993809, printed p.1, Epstein to Brachetti Peretti, 20 July 2014 18:01:32. Text layer. His reply of 18:22 — “ok / wait for you to know when you are in europe / thanks / Ugo” — is at https://epstein-data.com/EFTA00642520. ↩

  30. https://epstein-data.com/EFTA00642505, printed p.1, the same message in a different production copy — its date stands in a Date: field reading “Sun, 20 Jul 2014 17:23:33 +0000” against the Sent: field of “Sunday, July 20, 2014 5:24 PM” on the copy at the previous footnote — printing “could we meet in Paris tuesday night?” and “Ok i will organize to come in Paris in that week,let me know possible date to meet”. Text layer. The 11 July message stands alone at https://epstein-data.com/EFTA00993156. ↩

  31. https://epstein-data.com/EFTA00997586, of 7 October 2014 at 12:57:24, printed p.1; carrying Epstein’s 12:53 reply quoted beneath. Text layer. ↩

  32. https://epstein-data.com/EFTA00999597, of 20 October 2014 at 18:47:35, printed p.1; carrying the whole three-message exchange. Text layer. Epstein’s “27yth” also stands alone at https://epstein-data.com/EFTA02593736. ↩

  33. Falsification searches, run over the 158-document population by reading every page of every document’s text rather than by querying the index, on 11 September 2026. Each term in the list that follows was matched as a word-bounded form, without regard to case: fee matches the word fee and matches neither feel nor coffee nor a surname that contains those three letters. The one entry in the list that is not a word, 5%, was matched as a plain string; neither of the two contact lists it reaches carries it beside a quantity, and the plain-string rule is load-bearing for only one of them, the list on which the match falls inside a run of digits so that no word-bounded form can reach it. The two tests at the end were matched as word-initial forms. What substring matching returns instead is set out below, row by row and labelled as such, because four of these figures change under it. Terms searched, in the words a disconfirming document would use rather than in this page’s own vocabulary: commission, fee/fees, engagement letter, mandate, retainer (0 documents each); percent/per cent/5% (6 documents — https://epstein-data.com/EFTA00862961, https://epstein-data.com/EFTA00862963, https://epstein-data.com/EFTA00862965 and https://epstein-data.com/EFTA00862967, all copies of the 5 March 2015 thread, and the two contact lists https://epstein-data.com/EFTA00191467 and https://epstein-data.com/EFTA02779342, where in each case the extraction’s match stands in a telephone column and not in any sentence: on the first it falls inside a run of digits, which is the reason the plain-string rule is declared, and on the second it begins a line, where a word-bounded 5% reaches it); agreement/contract (6 documents, those four plus https://epstein-data.com/EFTA00863002 and https://epstein-data.com/EFTA00863004); invoice/wire/payment/paid (1 document, https://epstein-data.com/EFTA00191467, a contact list); signed/signature (2 documents — https://epstein-data.com/EFTA00191467 and https://epstein-data.com/EFTA00285104, where the word belongs to the name of an airport flight-services company in a contact list and a schedule); closed/completed/sold (2 documents, the September 2014 schedule copies). Under substring matching — which is not the rule above, and is set out here so that a reader who runs it is not left thinking the page is wrong — four of those ten rows change. fee/fees reaches 28 documents: 23 of them are copies of the scheduling message quoted in the text above, in which the letters fall inside “Or feel free to suggest a day/time”; the remaining five documents are the four contact lists and one schedule, and in them the letters fall inside “Coffee” on the schedule and in two of the lists, inside a surname in all four of the lists, and inside OCR damage in one of them. invoice/wire/payment/paid reaches 3 documents, the two further ones being a contact-list surname containing the letters wire. closed/completed/sold reaches 3 documents, the further one being a damaged word in a contact list. signed/signature reaches 3 documents, the further one being https://epstein-data.com/EFTA02641048, whose extraction carries the HTML class name gmail_signature: _ is a word character, so a word-bounded signature cannot reach inside it and a substring can. Not one of the extra matches is a fee, an invoice, a payment, a signature or a closing, which is why the word-bounded figures are the ones printed above and why the rule is declared: a reader who re-runs the test as a substring search gets 28 for fee, and this is where that 28 goes. Three figures in this footnote are corrections: an earlier draft of this page reported five documents for percent, 21 documents for the arab phrase and 12 passages for josh. The first two were low and the third does not re-derive in any unit; each is re-worked here. A fourth figure moved and has been moved back, and what moved it was the matching rule rather than the count: a later draft reported three documents for signed/signature, which is the substring figure, while the other nine rows carried word-bounded figures — so the rule this footnote declared described neither set. The rule is now declared and applied to all ten rows, the row stands at the word-bounded two, and the substring figure of three is recorded above with the reason it differs. The date range of the population was re-derived for this footnote rather than taken from a table of first-sheet dates. Every date token was extracted from every page of all 158 documents in each of the shapes this set uses — 19 Jun 2014, June 19, 2014, 6/19/2014, 2014-06-19, 11 luglio 2014, 23 gen 2018 and 2019/03/29 — and the minimum and maximum taken. The earliest date in the population is 19 June 2014; the latest is 30 March 2019, on the chat log at https://epstein-data.com/EFTA01617531. Four documents carry a date later than 19 June 2018: that log and the three messages of the 31 January 2019 thread left to another page (https://epstein-data.com/EFTA02628055, https://epstein-data.com/EFTA02628068, https://epstein-data.com/EFTA02628104). An earlier draft of this page asserted that no document in the population carried a date after 23 January 2019. That was false, and it was false because the test was run against a table of first-sheet dates instead of against the documents; the 2019/03/29 shape is invisible to every other date pattern in the list above. Year tokens of 2019 or later appear in 6 documents: those four, and the two contact lists https://epstein-data.com/EFTA00191467 and https://epstein-data.com/EFTA02779342, which carry 2020, 2021, 2024 and 2028. Two further tests were run the same way. arab* matches 25 occurrences on 25 pages in 23 documents. A substring search for the same four letters returns 34 occurrences, and every one of the nine it adds is the string arable inside “comparable” in the valuation memorandum — on https://epstein-data.com/EFTA00862961, https://epstein-data.com/EFTA00862963, https://epstein-data.com/EFTA00862965, https://epstein-data.com/EFTA00862967 and https://epstein-data.com/EFTA02508399 — and arable stands nowhere in the population as a word of its own, so 34 less 9 is the 25 printed here. Fifteen of those documents are copies of the single October 2014 chain, and in none of the fifteen does a name stand beside the phrase. Fourteen of the fifteen print “your arab friend” in their extractions; the fifteenth is https://epstein-data.com/EFTA01748376, whose extraction drops a letter from “friend” — the sheet itself reads “with your arab friend to talk about potential partnership”, read at 500 dpi — so a reader running the phrase against the index will be returned fourteen documents and not fifteen. The remaining ten occurrences are “Arabs” in the 5 March 2015 message, on the four copies named above, and “Arabia”, “Arabella” and a bare “Arab” across the four contact lists at https://epstein-data.com/DOJ-OGR-00022291, https://epstein-data.com/EFTA00191467, https://epstein-data.com/EFTA02779342 and https://epstein-data.com/EFTA02846578. josh* matches 22 occurrences on 7 pages in 7 documents, 16 of them the bare form josh; that test refuted an earlier draft of this page, which said the surname was not established: https://epstein-data.com/EFTA00843499 carries “Joshua Cooper Ramo” in its own To: field. These searches can refute the claims above; they were built from the disconfirming document’s vocabulary and not from this page’s, and the ones that did refute a claim were acted on rather than explained away. ↩ ↩2 ↩3 ↩4 ↩5

  34. https://epstein-data.com/EFTA00352468, of 25 February 2015 at 14:59:14, printed pp.1–2; sender field “Brachetti Peretti, Ugo”, carrying the same chain; and https://epstein-data.com/EFTA02084151, whose header block prints “Re: Jeffrey Epstein” over a Sent: of “Wed 2/25/2015 10:13:30 AM”. Text layer. ↩

  35. "w/ugo" returns 13 pages in 13 documents; all 13 are inside the 140 and all 13 were opened. Seven are schedules themselves — copies of Epstein’s day sheets covering the week of 1–8 March 2015 in Paris, except https://epstein-data.com/EFTA00285104, which is a whole-month March 2015 schedule of seven printed pages whose entry naming him stands on printed p.1. The other six are messages carrying the schedule block inside them rather than schedules: https://epstein-data.com/EFTA02391945 and https://epstein-data.com/EFTA02406240, of 25 February 2015, both carrying the sender field “Lesley Groff”; https://epstein-data.com/EFTA02084349, a two-word reply of the same day; https://epstein-data.com/EFTA02083597, of 26 February, whose sender field this extraction prints empty; and https://epstein-data.com/EFTA02511095 with its House Oversight copy https://epstein-data.com/HOUSE_OVERSIGHT_001615, of 4 March, both carrying the same sender field. In every one of the thirteen the entry naming him is an entry for Wednesday 4 March 2015, at 1:00pm, 1:30pm or 3:00pm. https://epstein-data.com/EFTA02391945 prints “1:00pm LUNCH Appt w/Ugo Brachetti Peretti” under the subject “1pm Lunch w/Ugo Wed. March 4”; https://epstein-data.com/EFTA00658625 prints “3:00pm Appt w/Ugo Brachetti Perett”; https://epstein-data.com/EFTA02511095 and https://epstein-data.com/EFTA00285104 print “1:30pm Appt w/Ugo Brachetti Peretti”. Text layer. The variation in the printed name across copies is OCR, not a difference between sheets. ↩ ↩2

  36. https://epstein-data.com/EFTA00352073, printed p.1, From: “Lesley Groff”, To: “Ugo Brachetti Peretti”, Date: “Tue, 03 Mar 2015 16:42:19 +0000”. Read at 400 dpi. It carries her 3 March 17:37 CET message, his “ok i will be there”, his 4:07 AM message of the same day and, beneath those, the 25 February instruction and the Paris address. The word before “meeting” in his 4:07 message is “te” on the image at 700 dpi of the parallel copy at https://epstein-data.com/EFTA00352069; the extraction of both sheets prints “to”. Names and telephone numbers elsewhere in that chain are barred on the image and are not supplied. ↩ ↩2

  37. https://epstein-data.com/EFTA02510071, printed p.1, From: “Lesley Groff”, Sent: “Tuesday, March 3, 2015 4:44 PM”, To: “Jeffrey Epstein”, subject “Ugo Brachetti Peretti”, the whole body being the quoted sentence. Text layer. ↩

  38. https://epstein-data.com/EFTA00352165, printed p.1, From: “Lesley Groff” over a barred address, To: “Ugo Brachetti Peretti” over a barred address, Date: “Wed, 04 Mar 2015 20:06:03 +0000”. Read at 400 dpi. It carries his 14:56 message quoted in full beneath it, and beneath that the 3 March 17:37 message with her name printed. The two spaces after the exclamation mark in her reply are on the sheet. ↩

  39. https://epstein-data.com/EFTA00654274, of 4 March 2015 at 11:02:12, printed p.1 of 1; from “Ugo Brachetti Peretti”, carrying Epstein’s 11:28 CET message quoted beneath. Read at 400 dpi. Epstein’s message also stands alone at https://epstein-data.com/EFTA00863094, timed 10:28 UTC. ↩ ↩2

  40. https://epstein-data.com/EFTA00639410, printed p.1 of 1. From: “Ugo Brachetti Peretti” over a barred address, Date: “Thu, 15 Oct 2015 15:39:24 +0000”, carrying Epstein’s 17:28 CET message quoted beneath. Read at 400 dpi and at 700 dpi for the last two words, which are “gas stattions” on the image against “gas stallions” in the extraction. ↩ ↩2

  41. https://epstein-data.com/EFTA01037648, of 23 January 2018 at 06:36:16, printed p.1 of 1; from “Ugo Brachetti Peretti”, carrying Epstein’s 07:02 CET message quoted beneath. Text layer. Three further copies of the same exchange are at https://epstein-data.com/EFTA01037649, https://epstein-data.com/EFTA02641048 and https://epstein-data.com/EFTA02641065; Epstein’s own message stands alone at https://epstein-data.com/EFTA02640993. ↩

  42. https://epstein-data.com/EFTA00853639, of 23 June 2015 at 21:11:18, printed p.1 of 2; from “Ugo Brachetti Peretti”, carrying the whole exchange quoted beneath: Epstein at 18:32 CET, the reply “No,how can i help?” at 20:02, Epstein at 22:16. Text layer. The individual messages also stand alone at https://epstein-data.com/EFTA00853594, https://epstein-data.com/EFTA00677954 and https://epstein-data.com/EFTA00853634. The link carried on the name is visible in the extraction of the parallel copies at https://epstein-data.com/EFTA02374675 and https://epstein-data.com/EFTA02495646; it points to a public modelling-gallery page and is not reproduced. ↩

  43. https://epstein-data.com/EFTA00853641, printed p.1 of 1, From: Epstein, Date: “Tue, 23 Jun 2015 22:19:27 +0000”. Read at 400 dpi. The To: field carries two bars, a short one where a display name would stand and a long one where an address would stand, so the recipient is withheld on the sheet and is not named here. The image prints “woody allen”; the extraction of this sheet prints “woody alien”. A second copy is at https://epstein-data.com/EFTA02497985. ↩

  44. https://epstein-data.com/EFTA02498012, printed p.1, From: wholly barred, Sent: “Tuesday, June 23, 2015 10:23 PM”, To: “jeffrey E.”; and https://epstein-data.com/EFTA02498179, printed p.1, From: Epstein, Sent: “Tuesday, June 23, 2015 10:45 PM”, To: a barred value. Both read at 400 dpi. The image of the second prints “friday july 3” in lower case throughout; the extraction of that sheet capitalises the month. This pair corrects an absence claim an earlier draft of this page carried, which said no answer to the 22:19 message was recorded; the answer is at these two identifiers and was found by running the absence test rather than by the search that built the page. ↩

  45. https://epstein-data.com/EFTA00659106, printed p.1, From: “TEODORANI Eduardo (CNH Industrial)”, Date: “Fri, 10 Apr 2015 10:33:28 +0000”, carrying Epstein’s message of 10 April 01:46 quoted beneath. Text layer. Epstein’s own message stands alone at https://epstein-data.com/EFTA00860105 and https://epstein-data.com/EFTA02505091; a further copy of the exchange is at https://epstein-data.com/EFTA02363723, whose extraction preserves the gallery link carried on the name. None of these four is in the "Brachetti" population. ↩

  46. https://epstein-data.com/EFTA00831277, printed p.1, From: Epstein, To: “Chiara Baschetti”, Date: “Sat, 19 Mar 2016 22:41:06 +0000”. Text layer. A second copy is at https://epstein-data.com/EFTA02468397. Not in the "Brachetti" population. "Baschetti" returns 15 documents across the corpus, all opened; the sender field of every one is Epstein, Brachetti Peretti or Teodorani-Fabbri, and none is hers. ↩ ↩2

  47. https://epstein-data.com/EFTA00627738, printed p.1 of 1. From: “Ugo Brachetti Peretti” over a barred address; Date: “Sat, 18 Feb 2017 05:09:16 +0000”; Epstein’s message of 17 February 21:09 CET is quoted beneath. Read at 400 dpi, where the name prints “fernando”; the extraction of this sheet prints “femando”, which is the rn→m fault. A second copy is at https://epstein-data.com/EFTA02338702 and Epstein’s own message stands alone at https://epstein-data.com/EFTA01053827 and https://epstein-data.com/EFTA02659547. ↩

  48. "Ferdinando Brachetti" returns 4 pages in 4 documents, all inside the 140: https://epstein-data.com/EFTA00453524, https://epstein-data.com/EFTA00453527, https://epstein-data.com/EFTA00499782 and https://epstein-data.com/EFTA02214888. All four were opened. All four belong to the June 2017 thread described in the text and none of them states a relationship between the two men. ↩

  49. https://epstein-data.com/EFTA00453524, of 6 June 2017 at 15:15:56, printed pp.1–2; from “Nicole Junkermann I Junkermanngroup” to Lesley Groff and Ferdinando Brachetti Peretti. Text layer. Cited here only to say where it is dealt with. ↩

  50. Outside record, cited as outside record and not as evidence for anything the documents are said to show. English-language Wikipedia articles on Gruppo API, on Ugo Brachetti Peretti and on Ferdinando Brachetti Peretti; Il Fatto Quotidiano, “Gli italiani negli Epstein file: Brachetti chiese aiuto per la cessione di Api”, 12 February 2026, and its follow-up of 14 April 2026 at https://www.ilfattoquotidiano.it/millennium/2026/04/14/ugo-brachetti-peretti/8344174/, whose visible text carries no statement from him, his representatives or the company; the article runs on past the visible preview behind the Millennium subscription wall and was not read in full for this page, so the finding is bounded to the text that was read and to the searches named next; Il Giornale d’Italia, 2026, on the same correspondence; and reporting of the SOCAR acquisition of Italiana Petroli of September 2025 and 2026. Searches run for a response: the subject’s name with “replica”, “smentita” and “nessun commento” alongside “Italiana Petroli” and 2026; the subject’s name with “Epstein” and the company name. No response was returned by any of them. Corpus searches generated from this record: "Italiana Petroli" returns 0 documents; "SOCAR" returns 3 pages in 2 documents, neither of which concerns this subject. ↩ ↩2

  51. https://epstein-data.com/DOJ-OGR-00022291 (printed p.7 of 95), https://epstein-data.com/EFTA00191467 (printed p.12 of 99), https://epstein-data.com/EFTA02779342 (printed p.13 of 50) and https://epstein-data.com/EFTA02846578 (printed p.7 of 95) are contact lists in which “Brachetti Peretti” is one alphabetical entry among many. Text layer. Nothing is taken from them beyond the fact of the entry; the neighbouring entries and the contact details printed in those lists are not reproduced. ↩

  52. Of the 140 documents, 137 carry an EFTA identifier, 2 a HOUSE_OVERSIGHT identifier (https://epstein-data.com/HOUSE_OVERSIGHT_001610 and https://epstein-data.com/HOUSE_OVERSIGHT_001615, both copies of Epstein’s Paris schedule) and 1 a DOJ-OGR identifier (https://epstein-data.com/DOJ-OGR-00022291, a 95-sheet contact list). Counted from the identifier string. ↩

Backlinks