This is the second note in a series updating release information in NARA’s JFK Database. The first note is here.
The JFK Database is a compilation of identification aids for the JFK Assassination Records Collection (ARC). It is an essential tool for working with the ARC. For details on both the database and the Collection, see the first note.
This second note looks at the details of how NARA released records in the Collection. It is based first of all on information in the latest update of the JFK database from 2021. Release information in the database is out of date by several years, so there has been much confusion on the status of records in the collection and how they were released.
This note will clear up much of that confusion by looking at the current status of Collection records released online by the National Archives (NARA).
Background
Here is a brief introduction to the JFK Act and the JFK Collection for people who don’t want to look up the first note (you should really look at it).
The JFK Collection was created by the JFK Act. The Act was signed into law by President George H.W. Bush (pere) on October 25, 1992. The main purpose of the Act was to gather as many government files related to the JFK assassination as possible, and to open the files to the public as quickly as possible.
To do this, the JFK Act designated the National Archives (NARA) as the repository for the collection, and created a limited term Assassination Records Review Board (ARRB) consisting of five board members (four historians and a lawyer). Together with a staff of about 30, the Board worked from 1994 to 1998, assembling a collection of over 300,000 records, of which 88% were fully available (open) to the public when the Board finally shut down in September 1998.
Information held back from public release was said to be “postponed.” The JFK Act mandated that “postponed” information would be released 25 years after the Act was signed, i.e., October 26, 2017.
From 1998 to 2017, records with postponements were mostly available to the public with passages held back. I call these redacted records. A very small percentage were not available to the public at all, except for some basic metadata. I call these withheld records.
The incomplete 1998 release
Why didn’t the ARRB finish releasing all the information in the ARC by 1998? The Board had only a limited amount of time in which to complete their tasks. They chose to spend a good part of their time looking for new records to add to the records collected by previous government inquiries: the Warren Commission, the HSCA, the Schweiker-Hart sub-committee established by the Church Committee, and so on.
The Board put aside records from these earlier investigations which they determined were largely, or entirely, irrelevant to the assassination. They did include these records in the JFK Collection, for the most part, ensuring their eventual release after October 2017. But they chose to spend much of their time looking for new information, and for records not previously reviewed or released.
They were also careful to make sure that truly relevant records, those central to the assassination, were reviewed in full and released as much as possible to the public.
The common claim that there was a lot of important information in the records the board put aside is false. A careful look at the discussions of records and passages released since 2017 shows exactly how little new and relevant information on the assassination has come out.
Although the releases since 2017 have not produced major revelations, the release of this information was mandated by the JFK Act, and it is worth a careful, critical look. In addition, these records have, in some cases, extended our understanding of the historical background of the JFK assassination. For historians, they remain a valuable source well worth looking at.
Releases from 2017 and later
If the JFK Act had simply stipulated that all information held back in the JFK Collection was to be unconditionally released on 26 October 2017, there would be very little to write about in this blog, and I would long ago have found another research interest.
But the Act did not do that. Instead, it allowed departments and agencies to request the President to continue postponing release of information in the JFK Collection, if “continued postponement is made necessary by an identifiable harm to the military defense, intelligence operations, law enforcement, or conduct of foreign relations” and if “the identifiable harm is of such gravity that it outweighs the public interest in disclosure.”
This part of the Act, section 5(g)(2)(D), was the basis for continued postponement of information in the JFK Collection for another 9 years, until 2026. In fact, there are still records where “postponed” information has not yet been released in full. Section 5(g)(2)(D) remains the legal basis for holding back this information from the public.
Section 5(g)(2)(D), or 5(g) for short, has made the release process for records in the JFK Collection very complicated. With 17 separate releases of records and information in the Collection since 2017, keeping track of the tens of thousands of pdfs NARA has posted online is a bean counter’s nightmare. This is certainly one reason there have been many false and/or exaggerated claims about the status of records in the collection.
This note gives a correct, detailed overview of these releases. If that’s not what you are interested in, skip this post. But if you want a more accurate, complete picture than what is available elsewhere, read on!
Release totals vs. record totals
As discussed in the first note in this series, NARA has posted pdf copies online of all the JFK records where information was previously held back.1They posted a few audio files as well. There have been 17 of these online releases, and a total of about 74,600 pdfs were posted. This is the release total.
However, many of these pdfs were copies of the same record. This is because the information held back in the records was released “incrementally”. Records with several words or passages held back often had one word or passage released on one date, and another word or passage released on another date. A release therefore does not equal a “record.”
How many distinct records were in these 17 releases? As we saw, there were about 37,226. This is the record total. This note will focus on record totals.
Record totals by year
A total of 37,226 records were released from 2017 to 2026. Remember that many of these records were posted by NARA multiple times, as different word and passages were released. Rather than ask just how many pdfs were posted each year (we saw that in the previous note), we might ask how many unique records were first posted in each year?
Here is what I found:
| year | uniq recs posted | pct of total |
|---|---|---|
| 2017 | 35434 | ~95% |
| 2018 | 1141 | ~3% |
| 2021 | 74 | <1% |
| 2022 | 473 | ~1% |
| 2025 | 94 | <1% |
| 2026 | 12 | <1% |
These numbers show that in 2017, NARA posted 35,434 of the overall total of 37,226 records. From 2018 to 2026, only 1792 more unique records were posted. That means over 95 percent of all records released were available at NARA by December 2017.
It does not mean that all these records were open in 2017! Many of these records were released with redactions, i.e. words and passage were still held back.
On the other hand, by December 2017, all “withheld” records eligible for release were released at least in part.2Eligible for release refers to information other than income tax returns, court sealed documents, grand jury records, and documents gifted to NARA that are restricted by a deed of gift.
Based on FOIA documents released by NARA in 2023-24, I believe this was a strategic decision by NARA. NARA was not able to unilaterally declassify records. It was, however, able to press reluctant departments and agencies to declassify as much material as possible, and to post all records with newly opened information as early as possible.
RIF records by release
As the numbers above show, by checking which records were posted and when, we are able to state when a pdf of any given record first became available online at NARA. We are not yet sure, however, when any given record was first released in full (RIF). Is there a way to check this?
Well yes, We can simply inspect every page of every pdf posted online at NARA. As the first note in this series pointed out, redacted material is replaced in these pdfs by an empty rectangular box. Find all the pdfs that have boxes and you know those records are not released in full. They are still redacted.
This is, of course, a vast task. There are 74,600 pdfs posted online at NARA, many of them dozens of pages long, some of them are hundreds, even thousands of pages long.
Is there a shortcut to avoid scanning the hundreds of thousands of pages in these electronic files? There are several. For instance, as noted above, many records are only one page long. These can be very quickly inspected for redaction boxes. I have done this several times, and I am confident all such records are now open.
Another heuristic is to look at records that were released only once from 2017 to 2026. Since the purpose of posting pdfs online is to show that the records are unredacted, a single post very likely indicates that this record was released in full on the very first go around, and there will be no redactions boxes in the record. I checked these records as well, and with about five exceptions that is indeed the case.
These two simple heuristics let us dispose of two thirds of the record total. In the end, of course, there is still a huge amount of extremely boring, time consuming scanning to do, but the task is not impossible. I did it, and I’ll stand by my record of reasonably accurate results for the following table.
The table includes the following basic release information: 1) release number, 2) release date, 3) total number of records released on that date, 3) records released in full (RIF) on that date, and the percentage of RIF releases versus the all releases on that date.
| # | rel date | total rel | RIF rel | pct RIF |
|---|---|---|---|---|
| #1 | 2017-07-24 | 3793 | 3788 | 99% |
| #2 | 2017-10-26 | 2891 | 2888 | 99% |
| #3 | 2017-11-03 | 676 | 96 | 14% |
| #4 | 2017-11-09 | 13213 | 0 | 0% |
| #5 | 2017-11-17 | 10744 | 8342 | 78% |
| #6 | 2017-12-15 | 4256 | 1780 | 42% |
| #7 | 2018-04-26 | 18897 | 5525 | 30% |
| #8 | 2021-12-15 | 1491 | 1483 | 99% |
| #9 | 2022-12-15 | 13235 | 8830 | 67% |
| #10 | 2023-04-13 | 422 | 422 | 100% |
| #11 | 2023-04-27 | 355 | 355 | 100% |
| #12 | 2023-05-11 | 502 | 501 | 99% |
| #13 | 2023-06-13 | 290 | 288 | 99% |
| #14 | 2023-06-27 | 1106 | 228 | 21% |
| #15 | 2023-08-24 | 21 | 5 | 24% |
| #16 | 2025-03-18 | 2565 | 2549 | 99% |
| #17 | 2026-01-30 | 139 | 137 | 98% |
This detailed breakdown can offer us insights into some of the more puzzling aspects of earlier releases, especially the 2017-2018 releases, which were quite confusing. Going into the details of this would make this post much too long, so I will put this off until my next note. Here we will just look at the overall picture.
First, however, there is an important caveat that needs to be mentioned. Remember that “open” records — records that have all the text/content opened to public — are only one of three record states in the ARC. Once a record is open, of course, redaction counters like me no longer follow it.
A record with text held back, i.e., a record that still has blank redaction boxes in it, can be a much more slippery item to keep track of. In records with multiple redactions on multiple pages, it often happens that some redactions are released, and some are not. This is not be reflected in the table above. If a record had even one redaction, I didn’t count it, even if hundreds of pages were newly opened on that date.
Why? The total number of redactions and releases in the collection is at least an order of magnitude greater than the total number of records with redactions; counting all of these exceeds my time and energy. All I counted were records where all the redactions were gone. The table is therefore an incomplete picture of how records were released. Readers should bear this in mind throughout the discussion.
From redacted to open: year by year counts
The total number of records released in full (RIF) for 2017-2026 was 37,216. Following the table above, we can track these records year by year. Here is a summary table:
| year | RIF recs posted/th> | pct of total |
|---|---|---|
| 2017 | 16894 | 45% |
| 2018 | 5525 | 15% |
| 2021 | 1483 | 4% |
| 2022 | 8830 | 24% |
| 2023 | 1799 | 5% |
| 2025 | 2549 | 7% |
| 2026 | 136 | <1% |
This summary table shows us how records were released under Trump 1.0, Biden, and Trump 2.0. Trump 1.0 was responsible for all the RIF releases in 2017-2018: this amounted to 60 percent of the total. Biden was responsible for all the RIF docs from 2021 to 2023, or about 33 percent of the total. The last two years were Trump 2.0, with 7 percent of the total records released in full.
This means that overall, Trump released about two thirds of the redacted records that were in the collection as of 2017, and Biden released one third.
These percentages differ from the vague, general discussions common online. If you have tried to follow these, you may well have come up with an impression which is the opposite of what actually happened.
Trump 1.0 has often been portrayed as releasing in full only a small fraction of the total number of redacted records. This is wrong. Trump 1.0 released the majority of the redacted material in the collection.
Biden’s releases have also been portrayed as minor, leaving most lengthy redacted passages untouched. Wrong again. Biden released in full almost one third of all redacted records in the collection. This fact has never been acknowledged online or in any of the numerous podcast discussions of releases.
Indeed, it was only after the 2022 releases that many writers at last conceded that there were no more records withheld in full in the collection. This was true by December 2017. NARA made a point of stating this on their website at the time. Online discussions ignored this, or even disputed it. Again, Wrong, wrong, wrong.
In contrast to the major releases of Trump 1.0 and Biden, the relatively minor 2025 releases of Trump 2.0 have often been described as larger than they actually were.
This is not entirely unwarranted. Trump 2.0 did release in full many records that had partial releases four or even five times before 2025. Why so many partial releases? Odd, I agree, but a careful look at what WAS released made it obvious that these redactions were not hiding assassination related material.
Based on oddness, however, many researchers convinced themselves that there was highly relevant information on the assassination in these postponed passages. Once more, wrong, wrong, wrong
I have written elsewhere about these records, which I examined carefully after release in full. I found almost all of them completely irrelevant to the JFK assassination. My own judgement is that both the number of passages released in 2025, and the significance of these releases, were exaggerated.
The most recent releases from January 2026 are trivial, in some cases simply providing clearer copies, rather than full releases.
My two cents
I should remind readers again that there are still records in the ARC that are not yet released in full. The most notable set, as I have said elsewhere many times, is the microfilm version of the CIA’s 201 file on Lee Harvey Oswald. Because this copy supposedly duplicates the hard copy version, which has been released in full, it has apparently been put on a back burner.
This is a mistake, I believe. In any case, the microfilm copy (which is over 30,000 pages long) is an integral part of the JFK Collection and should be released in full.
There are a handful of other records listed in the JFK database that have also not been released in full. The careful reader may have noted that I write of 37,226 records released and 37,216 records released in full. This does indeed indicate that there are at least 10 records that were posted at NARA from 2017 to 2026 which are still not released in full to the public. There are almost certainly a few others in addition to these, but that is a subject for another day.