What MKULTRA Actually Funded

The record survives because of a filing error

In January 1973, on his way out of the job, Richard Helms ordered the MKULTRA files destroyed, and they were. What is known about the programme today is almost entirely the contents of seven boxes that were not in the room when it happened.

The reason they were not is mundane. Financial papers for a sensitive project were normally kept by the branch that ran it, under the project file. In 1970, for reasons the CIA told the Senate are not known, the Budget and Fiscal Section sent these particular papers to the Retired Records Center as part of its own holdings. That put them outside the search in 1973, when the Agency went looking for MKULTRA material in order to destroy it, and outside the search in 1975, when it went looking in order to give it to the Church Committee. They turned up in 1977 because an archivist working through Freedom of Information requests read the listings of a section nobody had thought to check.

So the shape of the public record is an accident, and it is worth being precise about the shape, because it determines what can honestly be said.

A finance folder records a payment

Stansfield Turner, then Director of Central Intelligence, described the recovered material to a joint Senate hearing on 3 August 1977. The bulk of it was approvals for the advance of funds, vouchers and accountings, and his own assessment was that most of it was not very informative as to the nature of the activities it was supporting. Occasional project proposals are scattered through it, which is why a handful of subprojects can be described in any detail at all and most cannot.

What the folders did carry was names: 185 non-government researchers and assistants, and 80 institutions — 44 colleges or universities, 15 research foundations or chemical and pharmaceutical companies, 12 hospitals or clinics, and 3 penal institutions. Turner gave the names to the committee classified and withheld them from everyone else, citing the Privacy Act and the position of people who had taken money without knowing whose it was.

That last point is not incidental. Three of the subprojects were funding mechanisms whose entire purpose was to move money for the others. The Georgetown University Hospital file is the clearest surviving example of how they worked: a $125,000 grant passed through the Geschickter Fund for Medical Research, which the CIA's own memo calls a cut-out, arriving as a private donation and triggering a matching federal grant that doubled it. The memo is explicit about the point of the arrangement — the hospital was to be completely unwitting of the Agency's interest in it. Allen Dulles approved it in January 1955.

The column that is empty all the way down

The grid above asks each surviving body of evidence four questions. Who was paid, what the money bought, who it was done to, and what was found.

Read down the third column. Eight of the nine rows do not answer it. When the committee asked whether the people who had been given drugs without their knowledge could be identified and told, the Agency's answer was that no such names are part of these records, and it undertook to look for clues that might be enough to find them. None were.

This is not a gap that better archiving would have closed. A voucher records a transaction between the Agency and a researcher. The person the money was spent on is not a party to it and has no reason to appear. The one row that does name subjects is Orlikow v. United States, the case brought by nine Canadians who had been patients of Ewen Cameron at the Allan Memorial Institute in Montreal, and it names them because they brought it. It settled in 1988 for $750,000 with no admission; Canada later paid about eighty former patients C$100,000 each under a scheme of its own.

An institution's paperwork records what the institution did. It does not record who it was done to, and no amount of declassification changes that. The names exist in the public record only where the people they belong to put them there.

Fifteen categories, read into the record

Asked what the 149 recovered subprojects actually were, the CIA sorted them into fifteen categories and read the list aloud. It is the only enumeration of MKULTRA that exists and it is rarely quoted, which is understandable, because it does not support either of the familiar accounts.

The largest block, 56 subprojects, is research on behavioural drugs and alcohol. The second largest, 23, is motivational studies, assessments of defectors, and training techniques — work that would not look out of place in any intelligence service. Then eight on hypnosis, nine on sleep and behavioural change during psychotherapy, seven on simply acquiring chemicals, three on polygraphs. Four subprojects were on the magician's art, which the testimony glosses as sleight of hand useful for the surreptitious delivery of drug-related materials. Three were cancelled before any work was done. Three the Agency described as activities whose nature simply cannot be determined.

The counts that were given add up to between 137 and 141 of the 149. The rest sit in the two categories Turner read out without a number: library searches and conference attendance, and funding support for unspecified activities with the Army's Special Operations Division at Fort Detrick. A programme's own accounting of itself, produced under subpoena, does not reconcile to its own total.

Six

In that first category of 56, the Agency's breakdown was 17 subprojects probably involving no human testing, 14 definitely testing on volunteers, 19 probably testing on volunteers, and 6 testing on unwitting human beings.

Six is a smaller number than the account that treats MKULTRA as mass dosing of the American public. It is also a count of paperwork that survived by accident, compiled by the organisation being questioned, about a programme whose operational files it had destroyed four years earlier. Both of those are true, and neither cancels the other. The honest statement is that six is what the surviving financial record shows, and that the records which would have shown more were deliberately not kept in the first place — the 1963 Inspector General report noted that the practice was to maintain no records of the planning and approval of test programmes — and then destroyed.

That same 1963 report is what ended the activity in 1964. Its stated concerns were about exposure rather than harm: that the research would be considered professionally unethical, that some of it raised questions of legality, that a final phase of testing placed the rights and interests of U.S. citizens in jeopardy, and that disclosure could damage public opinion. The Church Committee, which found that the lives of subjects had been placed in jeopardy and their rights ignored, also recorded that when the covert testing was terminated none of those involved faced any disciplinary action.

One correction worth making while the dates are in view. MKULTRA ran from 1953 to 1964, not to 1973. The work continued under the name MKSEARCH until 1973, and 1973 is the year of the destruction order, which is probably how the two became one date.

The boxes that turned up this year

On 30 June 2026 the House Task Force on the Declassification of Federal Secrets held a hearing on MKULTRA. Its chair said that she and another member had been to Langley, and that the CIA was in the process of declassifying newly found documentation. She also said what it is about: a forgery programme that had been housed under MKULTRA.

If that holds, the new boxes will enlarge the category about documents and not the category about people. That is worth saying in advance, because the gap between "new MKULTRA files found" and what new MKULTRA files have historically contained is where most of the confusion about this subject is manufactured. The last row of the grid is marked unreleased for now and will be filled in when there is something to fill it with.

What none of this establishes

The surviving record cannot rule things out. It is a financial residue of an operation whose substantive files were burned, and an argument from its silence is weak in both directions — against the claim that something specific happened, and equally against the claim that nothing did. Where the record is silent here, the silence was arranged.

So: the claims that run furthest ahead of this material, the ones involving systematic programming of individuals across decades, have nothing in the released documents behind them. That is a statement about the documents. It is not the same as a finding that it did not happen, and anyone who tells you the files settle the question in either direction has not looked at what the files are.

What the record does establish is narrower and quite bad enough. A government agency paid 185 researchers at 80 institutions, most of whom did not know who was paying, to study ways of altering human behaviour; it tested on people who had not agreed to it; a man died; its own inspector general told it the work put citizens' rights in jeopardy; it stopped nine years before anyone outside asked; nobody was disciplined; and the files were destroyed by the man leaving the director's office. All of that is in the record because seven boxes were filed in the wrong place.