Chapter Nineteen

What the Files Say

Listen to this chapter, read by the author →

How I read the files

Before I show you anything I found in those files, you should know exactly how I read them, because it determines what I could and could not have found. I did not sit in the National Archives. I did not have a researcher. Everything in this chapter came out of a laptop, and the method is dull enough that you can check every step of it yourself.1, 2

Start with the thing that surprises people. The 2025 release is not a searchable database. It is a few thousand separate scanned files on a government website, one per record, named by record number, and every page is a photograph of a piece of paper. You cannot search them. You cannot even search one of them. Sixty years of secrecy ended in a format that is technically public and practically unreadable.1, 2

§So the first job was mechanical. Fetch every record in the release. Take the head pages of each one, run them through optical character recognition, keep the text with its layout intact, then throw the file away and move to the next. Sixteen agency series, a few thousand records, several nights of a machine grinding. At the end of it the collection is a folder of text you can search in a second, which is the thing the National Archives did not give anybody.1, 2

§Reading only the first pages of each record is a real compromise and I will defend it before I concede it. The cover sheet is where the machinery lives: originating agency, subject, classification, routing, and the opening criteria that say when a thing may be unsealed. Every discovery in this chapter came off a cover sheet. But the concession is total and I want it in plain sight. I have not read the bodies of most of these documents. If the interesting sentence is on page nine, I do not have it.1, 2

§With the text in hand the work stops being reading and becomes asking. And the questions that pay are never the exciting ones. Not who killed him. You search for a notation that keeps a page out of an index. For the designation an agency uses for a foreign liaison post. For the word that says a file cannot be opened without another agency’s permission. The story is not in the prose of these documents. It is in the stamps and the boxes and the routing.1, 2

And a search returns a great deal that is not there, which nobody tells you. Searching for the Australian security service by its initials returns thirty-nine hits, which for an argument that turns on Australia looks like a great deal. All thirty-nine are the middle of the word occasion. That is the actual work: not finding secrets, but establishing that a machine has misread a smudge, and checking every instance by eye before anything is said out loud.1, 2

So when I tell you in a minute that a particular notation appears twelve times and that eight of them came from one agency, that number is what is left after everything else was thrown out. It is a floor and not a total, it came off cover sheets and not full documents, and every instance was looked at by eye before it was counted. That is the standard the rest of this chapter is held to. If it turns out to be wrong, it will not be because I did not check. It will be because I checked the wrong thing.1, 2

Bureau Source 100

In the released FBI files there is a folder with an odd name on it. Bureau Source 100. In Bureau usage a source number is what you give an informant, a person outside the organization who is feeding you material. This one is not a mobster or a defector or a clerk in an embassy. Bureau Source 100 is a serving officer of the Central Intelligence Agency, and the file says who.3

The memorandum is dated the tenth of June 1953. It concerns James Angleton, who at that point ran a special unit inside the CIA’s foreign intelligence division and who, from the end of 1954, would run counterintelligence for the next twenty years. The Bureau writes that over the past year he has been very cooperative and has volunteered voluminous information. Then this: at the present time he more or less falls in category of Confidential Informant. So they recommend opening a control file on him.4

Ten months later, an internal memorandum from one Bureau official to another sets out how it works. Angleton, it says, frequently turns over information to the Bureau on a highly confidential basis. Then two sentences you should hear twice. In many instances the information is given to the Bureau without clearing through regular CIA channels. And in almost all instances it is given with the clear understanding that if it goes into anything leaving the Bureau, it will be written so that it cannot be traced back to either Angleton or CIA.5

Stop and look at what that describes. Material moving between two agencies outside the official channel, at the discretion of one man, on the standing condition that the finished paper must not be traceable to where it came from. That is not a conspiracy and nobody is hiding a murder in it. It is a working method, agreed between professionals, and written down in 1954 so that clerks would know how to handle it. Nine years before Dallas, this was already routine.6

One more line, and it is on the cover sheet rather than in the memo. When the review board processed these two documents it printed the condition for releasing them. Opening criteria: approval of CIA. An FBI memorandum, written by FBI men, about how the FBI handled a source, could not be opened to the public without the other agency agreeing. The rule that governs allied files governs this one too, and here you can watch it operate on American paper.7

Now the document that argues against me, and I would rather put it here than have somebody else find it. Sixth of January 1967. The FBI’s legal attache in London writes to the Director. A detective sergeant of the Criminal Intelligence Squad at New Scotland Yard has checked, and can find no record of Santo Trafficante entering the United Kingdom in the previous three years. Trafficante is one of the two men a Congressional committee would later name. The routing line says liaison, direct.8

Take that seriously, because it damages the lazy version of my argument. The line to a Five Eyes partner was open. It was routine enough that a request went out and a named police officer answered it inside a month, over a man’s passport stamps. Nobody had to invoke a treaty. So when I say allied files are closed on this, I cannot mean the channel was missing. It was there, it was warm, and people used it for small things constantly.9

†And that is exactly why it bothers me. A channel good enough to check whether a Florida mobster had been to Britain is a channel good enough to ask a partner service what it held on the murder of the President. It was pointed at travel records. Was it ever pointed at the thing itself, and if it was, where is the answer? I am not saying the friends were hiding something. I am saying the machinery was sitting right there and nobody has ever shown me it was used.2, 9

One last thing about how visible this is once you can search it. In the batch I indexed there are around fifteen folders whose title is simply CIA Liaison Material. Some carry a person’s name. Others carry a subject: Soviet Embassy Havana. Intelligence Matters Cuba. A student committee for travel to Cuba. Whatever else the relationship between those two agencies was, somebody in a registry had a label for it, a drawer for it, and enough of it to need a series.10

Bureau Source 6

When I found that the Bureau had given the CIA’s counterintelligence chief a source number, I went looking to see whether he was alone in that. He is not. There is at least one other, and the other one is not a person. Read across the released files and you keep meeting a phrase in brackets, in memoranda about Cuba, between 1960 and 1963. Bureau Source Six.11

Look at what it produces. In one document, Bureau Source Six reports, in a communication classified Top Secret, what the Cuban Foreign Ministry told the Czechoslovakian Embassy about a visa. In another it reports what a man in Rabat said to the Foreign Ministry of Cuba on a specific date in December 1962. In a third it reports what Cuban headquarters was told about a man in Mexico City. That is not an informant with good contacts. That is somebody reading the traffic.12, 13, 14

I would have left that as a strong guess, except for one line on a form. When the review board processed the Spellmeyer document it printed the condition for releasing it, and the condition is not the usual one. Not approval of CIA. Opening criteria: approval of N.S.A. That is the National Security Agency, whose entire function is the interception of communications, holding the key to an F.B.I. memorandum about Cuba.15

And every time it appears it comes with the same instruction, in the same voice as the one attached to Angleton nine years earlier. Information from Bureau Source Six should not be incorporated in any correspondence for dissemination outside the Bureau. It should be used for lead purposes only and must not be disseminated. In another file, plainer still: under no circumstances should Bureau Source Six data be set forth in reports.13, 14

Now the document that made me put my pen down. One of those memoranda, dated the twenty-eighth of January 1963, uses Bureau Source Six to report a conversation with the Cuban Foreign Ministry about a man’s salary being routed through an Algerian newspaper. And the subject of that memorandum, the man being tracked, is Richard Thomas Gibson. The national director of the Fair Play for Cuba Committee.16

Hold the dates together and be careful with them. In January 1963 the United States was reading Cuban diplomatic traffic and using it to follow the head of the Fair Play for Cuba Committee. That summer a young man in New Orleans wrote to that same committee, said he was opening a branch, was advised not to, and did it anyway. I am not saying one leads to the other. I am saying the organization he attached himself to was already inside a Top Secret collection effort before he touched it.13, 17

There is one more document worth knowing about and it reads like a stocktaking. A file headed CIA Liaison, listing what each service had given the other. Soviet mail drops abroad, handed over. Permission for the Agency to approach Soviet visitors inside the United States. Cuban secret writing techniques, identified by the Bureau and passed on. And this line: both F.B.I. and C.I.A. have had access to intercepted radio communications used by Cuban intelligence.18

So let me be exact, because this is mine and I would rather undersell it. What these documents establish is that in the years before Dallas the United States was reading Cuban diplomatic communications, routing the product into the Bureau under a source number, forbidding anybody to attribute it, and aiming it at the organization Oswald would later join. They do not establish that anything was heard about the assassination. They establish that the capability was live, it was pointed the right way, and its output was designed to be untraceable.2, 19

‡Look at the shape one last time. A collection capability shared between services. A number instead of a name. A standing instruction that the product must never be traceable to where it came from. And a release, sixty-two years later, gated on the approval of the agency that did the collecting. That is the machinery I described three chapters ago, and I did not have to theorise a word of it. I found it by reading four documents nobody had indexed.2

The Foreign Liaison Unit

Going back through the index I built, I found the memorandum about the Fair Play for Cuba Committee’s director a second time, in a different file, as a different copy. The message is the same. The copy is not. This one is stamped Top Secret, and around the edges of it, in the routing and the stamps that nobody reads, is a set of things I have spent this whole book describing separately. They are all on one page.20

Look at who got a copy. One to the legal attache in Paris, personal attention. One to the legal attache in Madrid, personal attention. Those are the Bureau’s men inside American embassies abroad, and they are the same two posts that turn up elsewhere in this book, handing a letter to a Controlled American Source and asking Franco’s military intelligence to find a man. Top Secret intercept material on an American political organization, sent to Bureau officers in two European capitals.21, 22

And then, in the routing block, four words I had not seen before in this material. Foreign Liaison Unit. Not a euphemism, not a bracket, not a phrase somebody invented to avoid naming an agency. An organizational unit, with a name, on a form, whose function was moving this kind of paper between the Bureau and services abroad. I have spent three chapters describing that machinery from its outputs. Here it is with a desk and a job title.23

And at the foot of the page, handwritten, the notation I taught you to look for in the chapter on how paper moves. Not recorded. So: Top Secret intercept product, about the head of the organization Oswald would join that summer, sent to Bureau officers in two foreign capitals through a unit that exists to move exactly this, and deliberately kept out of the central index so that no future search would ever surface it. Every one of those is a separate decision.21

I want to be careful, because a document that confirms everything you already believe is the one to distrust hardest. It does not say anything about the assassination. Nobody in it is planning anything. What it does is put five things I had been describing separately onto one sheet of paper: signals intelligence with a source number, an American political group as a target, a foreign liaison channel with a name, two overseas posts, and a decision not to file. That is the plumbing, photographed.2

†One more from the same file, and I am going to report it and then refuse to do the obvious thing with it. A 1972 memorandum from the Agency to the Bureau about a man referred to only by a cryptonym, coming to New York for what it calls special training, and offering something. We can also have him perform any appropriate tasks which your Bureau may desire to be done. Replies to go by a named channel. It sits in the file of the Fair Play for Cuba Committee’s director.24

Every instinct says that cryptonym belongs to the man whose file it is in, and that would mean the head of the organization Oswald joined was a source being offered around between two agencies. I cannot show that. The name is not on the page I can read, files hold material about more than one person, and I have spent a whole book telling you what happens when somebody closes a gap like that with a feeling. So there it is, with its number, unclosed. Somebody with better copies can finish it.2, 25

The agency nobody mentions

Sixty years of argument about this case is conducted almost entirely between three initials. The Bureau, the Agency, the Service. Ask anybody who follows this and they will give you a view on all three. Almost nobody mentions the fourth, which in 1963 was larger than any of them, collected more than all of them combined, and has never been asked a single public question about the assassination of the President.2

Here is a number I have not seen anybody publish, and you can check it yourself. The March 2025 release contains three hundred and fourteen Bureau files. I indexed them, and twenty-seven carry the same line on the National Archives release form. Opening criteria: approval of N.S.A. Not the Agency. Not the Bureau that wrote them. Better than one file in twelve of the F.B.I. paperwork in the President’s assassination collection could not be opened without the permission of the National Security Agency.26, 27

And then I opened them. Some are not summaries at all. One is headed with a codeword and states, in its own first paragraph, that the information in it was derived from Latin American communications, and gives the message, the date, the system and the worksheet number for each item. That is not a document about signals intelligence. That is signals intelligence, in the President’s assassination collection, with its filing references still attached.28

One of them is a message. From El Paso, Texas, on the twenty-first of August 1961, addressed to the Prime Minister in Havana, to Fidel Castro personally, naming an underground plot and the exile leaders behind it. Somebody read that on its way, decrypted it, wrote it up, and filed it. Sixty-four years later it is in a collection about the murder of an American president, and it took the National Security Agency’s approval to let you see it.29

Across the head of that document, in capitals, is the handling rule. Special handling required. Not releasable to foreign nationals. The information contained in this document will not be disclosed to foreign nationals or their representatives without the express approval of the Director, National Security Agency. That is the machinery I have been describing for three chapters, printed on the page, naming the man who holds the veto.30

And now I have to turn that around on myself, because it cuts hard. Not releasable to foreign nationals means exactly what it says, and in that period it meant the allies too. So the most sensitive American material on Cuba was specifically marked to be kept away from London, Ottawa, Canberra and Wellington. Which weakens one version of my argument considerably. If the best of it was never shared, the friends may genuinely have had less than I have been assuming.2

‡Except it does not answer the question I have actually been asking. What America withheld tells you nothing about what the partners collected themselves. Britain had its own stations, its own targets and its own product, and the whole point of the arrangement was that both sides were listening independently and exchanging the results. So the caveat narrows my claim to this: I am not asking what Washington told them. I am asking what they heard.2

There is one practical consequence. My letter to that committee already asked the National Security Agency a general question, and a general question is easy to answer with nothing. So I have rewritten it around the count. Not what do you hold, which invites a shrug, but: name every record in this collection where your concurrence was required before release, say why, and say when each was last looked at. That is a list they either have or do not have.2

§Then it occurred to me that I had asked a narrow question. I had searched for one agency’s name in that field because I already suspected it. So I went back and pulled every value that appears in the opening criteria box across the whole release, and read the list. It is not one agency holding keys to this collection. It is nine.31, 1

§Here is the league table, and it is a floor rather than a total, because a machine read it. The Central Intelligence Agency gates about eighty files. The National Security Agency gates about twenty-six. And in third place, closer to the N.S.A. than the N.S.A. is to the Agency, sits the Internal Revenue Service, with about twenty-one.31, 1

§After that it becomes a queue of single files, and the list is worth reading aloud. The State Department holds one. The Drug Enforcement Administration holds one. Immigration holds one. The Bureau of Prisons holds one. The United States Coast Guard holds one. And one document in the collection cannot be opened until it has been reviewed by the Adjutant General of the Army.31, 1

No committee sat down and arranged that. What you are looking at is the residue of a review conducted file by file in the 1990s, in which every agency that had ever touched a page got to write its own name in the box, and nobody afterwards stood back and looked at the resulting list. A hundred and thirty-one documents in the President’s assassination collection are held shut by nine different bodies, and no single person or office has authority over all nine.31, 1

And that is a stronger point than the one I set out to make. I had been telling you to look at the National Security Agency, because signals intelligence is where my argument lives. The truer thing is duller and worse. There is no gatekeeper to petition. There are nine, they do not answer to each other, the independent body that could have overruled all of them expired in 1998, and one of them is the tax authority.31, 1

Bern, and a name I will not print

Read enough of these and you keep hitting the phrase: the other Government agency. It is deliberate vagueness, so a document can move without naming who supplied it. In a 1964 cable from Bern, somebody spells it out. The other Government agency, it says, is the Research Branch, and then, in brackets, covert CIA. At the Embassy in Vienna. The euphemism, translated by the people using it.32

Somebody in Europe wrote a letter about the assassination. It reached an American embassy, was translated, was written up as a two-page insert, and twenty-five copies were sent to the FBI’s Dallas office, with fifty more enclosures counted on the distribution list. One letter from one stranger, and a government photocopied it seventy-five times. That is what it looked like when a state could not afford to ignore anything.33

Here is the FBI doing something careful, in 1964, that most of the internet still will not do. The cable notes that material from Bureau files about a man of the same name was deliberately left out of the report, because, it says, identity has not been proven and the information is not pertinent in any case. Somebody chose not to attach a file to a person on the strength of a matching name. That is not a small thing.34

Somebody has written two words by hand on the Bern cable, next to the file number: not recorded. It means the document was moving through the system without being entered in the central index. Findable only by whoever already knew where it was. Multiply that by an archive. A great deal of what was hard to find for sixty years was not hidden. It was just never written down in the place you would look.35

The Bern cable names the man who wrote the letter. I am not going to. The Bureau itself concluded his identity had never been proven, and declined to attach its own file to him on the strength of a matching name. Repeating it now, to an audience of millions, would undo a piece of care somebody took in 1964. Released is not the same as fair game, and that distinction is going to come up a lot.36

Counting the invisible files

Three times in this book I have shown you a document with two words written beside the file number. Not recorded. It means the page moved through the Bureau without ever entering the central index, so no future search would surface it. Each time I treated it as a curiosity worth pointing at. Then it occurred to me that pointing at three of something is not evidence of anything. Counting all of them might be.2

§The 2025 release runs to thousands of scanned pages, and scans are pictures, not text. So I ran the whole pile through optical character recognition and searched the output for the notation in every mangled form the software might produce. Which brings the honest problem straight away. Not recorded is a rubber stamp, sometimes a handwritten scrawl, sometimes sideways in a margin. Machines read printed text well and stamps badly. Whatever number comes out of this is a floor, not a total.2, 1

§The first result is an absence, and absences are usually the most reliable thing a search gives you. Across more than a thousand Agency documents, that handwritten notation appears not once. Every instance is Bureau paper. Which invites an obvious conclusion, that this was the habit of one building in Washington. So I went and searched the Agency material for the idea rather than for the words.37, 1

§The Bureau wrote not recorded by hand, in the margin, one document at a time, and I found twelve. The Agency did not need to write anything. On the standard cable form there is a row of little boxes for what should happen to the page, and one of those printed boxes says no index. I have found it ticked or marked on thirty-eight of them. So it was never a quirk of one building. One of them improvised it and the other one had it built into the paper.38, 1

Three of the eight sit next to each other in the same file, consecutive record numbers, all three top secret intercept product. And the name the Bureau filed them under is Eusebio Azcue: the Cuban consul in Mexico City who argued with Oswald across a counter in September 1963 and later said the man he saw was not the man who died in Dallas. Three unindexed signals intelligence documents, filed under him.39

Because I read them, and you should know what is actually in them. They are dated 1967, four years after Dallas, and they are not about Oswald at all. One is intercepted Brazilian diplomatic traffic listing Colombian communists expected at a conference in Havana. That is what is on the page. Filed under the consul’s name, yes, but the contents have nothing to do with the assassination, and if I let you keep the version from thirty seconds ago I would be doing the thing this book exists to argue against.40

One document in the twelve I cannot explain away. Dated the twenty-seventh of December 1963, five weeks after the assassination. It originates with the National Security Agency. It is top secret. Beside the file number, not recorded. Down the right hand margin runs the routing slip of the Bureau’s entire executive floor: Tolson, Belmont, Mohr, DeLoach, Conrad, the men directly under Hoover. And in the box marked opening criteria, the archive has typed one word before the agency’s name. Indefinite.41

Look at what that is, layer by layer, because each layer is a person choosing. Somebody at Fort Meade decided this belonged with the Bureau. Somebody at the Bureau decided the whole executive floor should see it. Somebody with a pen decided it would not go in the index. And a reviewer in 1998, thirty-five years later, decided there was no date on which it could open. Four decisions, four people, one page, and the last of them is still in force today.41, 2

§So let me be exact about the claim. These are floors from an imperfect machine read of a fraction of the pile, and I would expect both numbers to rise. I do not know what most of these documents say. I cannot show that any page was kept out of the index because of the assassination rather than because that is simply how secret traffic was handled. What I can show is that both agencies had a routine way of moving a document so that no future search would find it, and that in the Bureau’s case it falls overwhelmingly on the one kind of material that still cannot be opened without permission.2

Which is why I am printing the record numbers rather than the conclusion. A machine found twelve; a person with the patience to turn pages would find more, and somebody ought to build the list properly, because a catalog of the documents deliberately kept out of the catalog is exactly the reference tool this case has never had. I have done a fifth of it badly. I would rather hand you that than a tidy answer I cannot support.2

† disputed   ‡ theory   § reconstructed — everything unmarked is established. Superscript numbers key to the sources below. Corrections

Sources for this chapter

41 sources. Every claim above is keyed to one of them. The complete list for the whole book runs to 1,276.

  1. the reconstruction
  2. the author's assessment
  3. FBI HQ file 62-99724, RIF 124-10326-10098 and 124-10326-10100, released 2025
  4. FBI memorandum, 10 June 1953, subject Information Received from James Angleton, RIF 124-10326-10098
  5. FBI memorandum, Branigan to Belmont, 30 April 1954, subject James Angleton, RIF 124-10326-10100
  6. RIF 124-10326-10098 and 124-10326-10100
  7. JFK Assassination System identification forms for RIF 124-10326-10098 and 124-10326-10100
  8. FBI memorandum, Legat London to Director, 6 January 1967, file 92-2781, RIF 124-10204-10215
  9. RIF 124-10204-10215
  10. FBI record series 124-90137 through 124-90140, released 2025
  11. FBI records in the 2025 release, RIF 124-90137-10052, 124-90146-10081, 124-10291-10123
  12. RIF 124-90137-10052
  13. RIF 124-90146-10081
  14. RIF 124-10291-10123
  15. JFK Assassination System identification form for RIF 124-90137-10052
  16. RIF 124-90146-10081, FBI memorandum, 28 January 1963, subject Richard Thomas Gibson
  17. Warren Commission Hearings, Oswald's correspondence with the Fair Play for Cuba Committee
  18. RIF 124-10325-10318, FBI memorandum on CIA liaison
  19. RIF 124-90137-10052, 124-90146-10081, 124-10291-10123, 124-10325-10318
  20. RIF 124-90120-10036, FBI memorandum, 28 February 1963, classified Top Secret
  21. RIF 124-90120-10036
  22. compare RIF 124-10114-10010 and the Madrid referral discussed earlier
  23. RIF 124-90120-10036, routing block
  24. RIF 124-90146-10072, CIA memorandum to the FBI, 15 June 1972
  25. RIF 124-90146-10072
  26. JFK Assassination System identification forms, 2025 release, FBI series
  27. the author's index of 317 documents
  28. RIF 124-90084-10090, classified Top Secret with COMINT codeword
  29. RIF 124-90055-10019, decrypted message traffic, 21 August 1961
  30. RIF 124-90055-10019, handling caveat
  31. 2025 release, National Archives identification forms, opening criteria field
  32. FBI AIRTEL, Legat Bern to Director, 21 August 1964, classified SECRET
  33. FBI AIRTEL, 21 August 1964, enclosure and distribution counts
  34. FBI AIRTEL, 21 August 1964, Bureau file material excluded on the ground that identity was unproven
  35. FBI AIRTEL, 21 August 1964, bearing a handwritten 'NOT RECORDED' notation beside the file number
  36. FBI AIRTEL, 21 August 1964, name withheld here by editorial choice
  37. 2025 release, FBI 124- and CIA 104- series
  38. 2025 release, FBI 124- series
  39. 2025 release, RIF 124-90039-10003, -10004, -10005
  40. 2025 release, RIF 124-90039-10005
  41. 2025 release, RIF 124-10299-10044