The men who questioned Hans Bühler kept telling him that his employer was a spy center. He thought they were wrong, and he said so, for up to five hours a day.
He was 51, a salesman of encryption machines for Crypto AG in the Swiss canton of Zug, and the trip he made in March 1992 was his twenty-fifth to Iran. On 18 March, two days before he was due to fly home, Iranian intelligence agents detained him and took him to a military prison in Tehran. The grounds, as the Swiss embassy reported them, were spying, bribery, illegal contacts and the illegal use of alcohol. The alcohol, in his account, was a cola with homemade vodka at a private flat. The bribery was money for a sick child's medicine and a taxi.
What he could see of his interrogators was their shoes. According to the account the Swiss journalist Res Strehle reconstructed with him two years later, he was kept blindfolded during questioning, in a cell about two meters by three, with a felt mat, two wool blankets, two plastic buckets and a strip of cardboard he could push under the door when he needed something. The rules on the wall, in Farsi, Arabic and English, required absolute silence. "I was never beaten," he told the Baltimore Sun, "but I was strapped to wooden benches and told I would be beaten." When Swiss consular officials first saw him, on 20 April, they found him in bad mental condition. On a later visit, about two months in, a Swiss diplomat was allowed four minutes with him, in English, and not a word about the case.
The suspicion his interrogators kept putting to him was correct, and Bühler had no way of knowing it. Since June 1970 Crypto AG had been secretly owned by the Central Intelligence Agency and West Germany's Bundesnachrichtendienst. Many of its machines, sold to more than 120 countries, were built so that the two agencies could read what those governments believed they were hiding.
That much has been public since February 2020, when the Washington Post, the German broadcaster ZDF and the Swiss broadcaster SRF reported what the Post called "the intelligence coup of the century." It is usually told as a story about a secret that held for half a century and then broke. The record tells it differently. The allegation was on Swiss television in 1994, in an American newspaper in 1995 and in Der Spiegel in 1996. Iran, whose interrogators had accused the company of exactly this, went back to buying the machines almost immediately. And in Switzerland the answer and the question sat in two offices of the same state. In May 1995 the federal police closed an inquiry with the finding that the true owners of Crypto AG could not be identified. The country's foreign intelligence service had known since the autumn of 1993, and, a parliamentary inquiry later found, it told the police it had no indication that foreign services stood behind the company.
What held was narrower: the fullest account stayed with the people who did it, and the people asking never got to see it.
A Handshake at the Cosmos Club, 1951
Crypto AG began as the life's work of Boris Hagelin, a Swedish engineer whose M-209 cipher machines the United States Army bought by the tens of thousands in the Second World War. After the war he moved his business to Switzerland, where neutrality was a selling point no American or German competitor could match. In 1951, according to the CIA's internal history as reported by the Post, Hagelin and William Friedman, the founding father of American cryptology, reached a handshake agreement over dinner at the Cosmos Club in Washington: Hagelin would sell his most secure models only to countries the United States approved, and older, weaker systems to everyone else. He received up to $700,000 up front, and a licensing deal followed in 1960.
The arrangement rested on one man: his loyalty, his discretion, and his staying alive. By the late 1960s Hagelin was nearing eighty, and the question of who would run the company after him had no answer the Americans trusted. The two intelligence services removed the dependency on a person by replacing it with a structure. They bought the company.
Bearer Shares in Liechtenstein, June 1970
The purchase was arranged through a Liechtenstein law firm, Marxer and Goop, using shell entities and bearer shares that required no names in registration documents. The CIA and the BND each took half. The Post, from the CIA history, puts the price at about $5.75 million; the Crypto Museum, from other documents, at 25 million Swiss francs in three installments. The operation was named Thesaurus and later Rubicon, and inside the CIA the company was called Minerva. Siemens provided cover and advice for five percent of sales, the NSA shaped the algorithms, and by 1981 the largest customers were Saudi Arabia, Iran, Italy, Indonesia, Iraq, Libya, Jordan and South Korea.
The design principle that made the business work was a set of weaknesses in the algorithms that only someone who knew how they were built could exploit. Approved countries got strong versions; the rest got versions the NSA could read. A customer who suspected nothing had no reason to compare.
Everything about the structure was designed so that a question would arrive nowhere. A journalist asking who owned Crypto AG would reach a Liechtenstein foundation. A customer asking whether the algorithms were sound would reach an engineer who, in most cases, believed they were. A salesman asked whether his company worked for a foreign intelligence service would say no and mean it.
The Irreplaceable Man, Munich 1979
A rigged algorithm needs someone inside the company who is allowed to decide what the algorithm is. From 1979 until he left the company in 1994, that person was a Swedish mathematician named Kjell-Ove Widman. According to the CIA history as reported by the Post, he was recruited in Munich in 1979, became the company's scientific adviser, and was regarded by the agency as its "irreplaceable man." The purchase of 1970 had removed the dependency on one man. Nine years later the agency had another.
Engineers at Crypto AG built the machines, tested them and sold them, and most of them, by the accounts that have since emerged, did not know what their employer was. The cipher logic, the part that decided whether a machine was strong or readable, came down to a small number of decisions that ran through one office. The Baltimore Sun reported in 1995 that Widman had total authority over the company's algorithms, and quoted a longtime colleague who said that Widman would often travel to Germany and return with design instructions. Widman denied that anyone outside the company controlled the work.
According to the BND's own history, the partners set a rule the weaknesses had to obey. They had to be "undetectable by usual statistical tests" and, if discovered, "easily masked as implementation or human errors." The flaw was designed to look like the kind of mistake every commercial product contains, so that finding it would prove nothing.
Hundreds of employees could testify honestly that they had never seen anything improper, because the improper part never passed through their hands. A customer's own cryptographers could study a machine and find only what looked like an ordinary error, while the one person who could say otherwise had every reason not to. At each level the arrangement put the knowledge in as few hands as possible and the plausible innocence in as many as it could find. Secrecy protects an operation from the people who know. This arrangement went further and recruited the people who did not.
What 40 Percent Looks Like
At times, including in the 1980s, Crypto AG machines accounted for roughly 40 percent of the foreign government communications that NSA cryptanalysts decoded, according to the documents. That is a share of what the NSA managed to break, not of everything governments sent. During the Iranian hostage crisis of 1979, the former NSA director Bobby Ray Inman later said, the agency could answer President Jimmy Carter's questions about Iran's reactions "about 85 percent of the time." Argentina's traffic in the Falklands War was passed to Britain. The evidence President Ronald Reagan cited against Libya after the La Belle discotheque bombing in 1986 came from decrypted Libyan traffic. In the war with Iraq, the CIA history describes Iran's traffic as "80 to 90 percent readable."
The Soviet Union and China never bought from Crypto AG. Their suspicion of a Western company protected them: the governments most likely to assume they were being watched were the ones this channel could not reach. The machines worked on governments that trusted Swiss neutrality enough to stake their diplomatic secrets on it.
In March 1992, a program that delivered a substantial share of what the United States could read about the world's governments depended, briefly, on whether a man who knew nothing would say something.
Hydra, 1992
The CIA called the Bühler case Hydra and later described it in its own history as "the most serious security breach in the history of the program." Why Iran took him is less clear than it first looks. The formal charges did not mention rigged machines, and according to the Crypto Museum's reading of the CIA history, the agency later concluded that he had probably been taken as leverage for an Iranian who had recently been arrested in Switzerland. The accusation in the interrogation room was real. Whether it was the reason for the arrest, the record does not settle.
Getting him out required money that no one could be seen to pay. In September 1992 Iran set bail at one million dollars. The CIA, citing its policy against paying ransom, refused to contribute. Crypto AG paid, and according to both the Post and the Crypto Museum, the money was secretly supplied by the BND. Bühler came home to Zurich on 5 January 1993, after nine and a half months, most of it in solitary confinement.
The company fired him on 10 March 1993. At one point he said it had asked him to repay the bail, a claim he later withdrew. In December 1993, according to the Crypto Museum's account, he signed a first settlement with the company, for 250,000 Swiss francs, in which he agreed not to press for more money. It said nothing about talking to the press. Bühler, who had believed throughout his detention that the accusations were absurd, began to doubt it. In March 1994 he said so on Swiss and Austrian television, and Res Strehle's book about him, "Verschlüsselt," appeared the same year.
The Lawsuit That Closed the Courtroom
What happened next is where the private doubt nearly became a public record, and where, according to the CIA's own account, it was steered away. That account comes to us second-hand: the CIA history has not been published, and the Crypto Museum's reconstruction of its passages on the case is the most detailed available. That the case was settled, and when, is confirmed independently by the Sun and Der Spiegel; the strategy and the terms are not.
The company and its secret owners faced a choice. Settling quietly with Bühler would have looked to customers like an admission of guilt, while suing him was more credible but risked exposing the operation. In my reading, the lawsuit was the company's denial in its most persuasive form: a firm that takes its accuser to court looks like a firm with nothing to hide.
In July 1994 Crypto AG's lawyers obtained a temporary restraining order against Bühler, pending a lawsuit against him. The agency decided that August to let the case proceed and to draw it out for years, in the hope that Bühler would run out of money or energy and settle. In July 1995 the court ruled that both sides had to prove their points, which meant the company would have had to show that its machines had never been manipulated. In late September, in breach of the restraining order, Bühler gave a long interview to the Baltimore Sun. For the agency this was an unexpected break: the breach opened the door to an out-of-court settlement. Weeks later the case was settled, days before former Crypto engineers were due to testify that they believed the machines had been altered, the Sun reported. The parties agreed not to disclose the terms.
The second settlement closed the gap the first had left. Its terms were the whole mechanism in miniature. Bühler had to stay away from the press, at a cost of 100,000 Swiss francs for every violation. He had to declare that there were no grounds for his accusations. The settlement put a price on his voice. The company paid his legal bills up to a cap. Der Spiegel reported in 1996 that he had since kept an iron silence about the case.
Testimony from inside the factory, entered in a court file, would have been a different kind of document from anything a reporter could write. The court never heard it. Two of the people the Swiss federal police questioned in 1994 described alleged manipulation; their statements stayed in police files. The interview that put Bühler's story into an American newspaper came weeks before the settlement that took it out of a Swiss courtroom. In those same months the federal police in Bern were trying to answer the question the lawsuit had raised, and another office of the Swiss state already knew the answer.
The Story Was in the Baltimore Sun in 1995
On 10 December 1995, as part of a series on the National Security Agency, the Baltimore Sun's Scott Shane and Tom Bowman reported evidence that Crypto AG machines had been rigged for American codebreakers. Bühler, in 13 years with the company, had "no inkling." Former engineers said they believed outside parties had shaped the algorithms. A confidential memo from 1975 listed among the participants in a design meeting a woman named "Nora Mackabee," whom a former engineer identified as an NSA cryptographer. Asked by the Sun about her work with Crypto, she said: "I can't say anything about that."
The article could show a pattern, a memo, a name and the testimony of people who had worked inside. It could not show the ownership, because the ownership sat behind bearer shares in Liechtenstein. The reporters named the missing piece themselves: "The whole story will be told only when secret U.S. documents are declassified, probably well into the next century." Crypto AG had already written to its customers in 1994 that "manipulation of Crypto AG equipment is absolutely excluded," and it dismissed the allegations as the invention of disgruntled former employees. In early September 1996, Der Spiegel published its own account, "Wer ist der befugte Vierte?", which linked the company to the NSA and to Germany's cipher authority. The company denied that too.
By the end of 1996, anyone who wanted to believe the machines were compromised had grounds. Anyone who needed to keep using them had a denial, and nothing that could force the company to retract it.
The record cannot tell me how many governments read those articles and quietly moved their most sensitive traffic elsewhere; that kind of exit leaves no cancelled contract behind.
Iran Bought Again
What the CIA history does record is how the customers behaved. The publicity around Hydra cost the program: at least half a dozen countries, including Argentina, Italy, Saudi Arabia, Egypt and Indonesia, cancelled or suspended their contracts. The chief executive, Michael Grupe, went on Swiss television and disputed what the Post, drawing on the documents, called charges he knew to be factual. "Grupe's performance was credible, and may have saved the program," the CIA history says.
Iran did not cancel. The country whose interrogators had accused the salesman's employer of exactly what was true "resumed its purchase of CAG equipment almost immediately," according to the CIA's own account. No document I have found explains Iran's decision, and this piece will not invent a motive. The plausible explanations are all ordinary. A government under sanctions had few suppliers to choose from. Replacing a national cipher system takes years and money, and the machines already installed did not stop working on the day a suspicion was voiced. A suspicion voiced in an interrogation room gave a procurement office nothing it could act on.
The dependency ran in both directions. The NSA needed the machines in order to read. The customers needed the same machines in order not to be read. Both sides relied on one piece of infrastructure, and only one of them knew what it did. That is why Grupe's denials mattered so much: the agency depended on the customers' trust as much as the customers depended on its product. The question this leaves is larger than one company: when a country depends on a system someone else built, who, apart from the builders, can establish what it actually does?
The Germans read the situation differently. Hydra had rattled the BND, and after reunification the risk looked larger than the reward; in September 1993 the CIA agreed to buy out the German share for $17 million, and by mid-1994 it was the sole owner. The Americans kept going through the arrest, the bail and the television interviews. The Bühler affair, the CIA history says, did not cause the program's demise, and "at the turn of the century Minerva was still alive and well."
But by the mid-1990s, the same history says, the days of profit were long past and the company survived on American government money, while encryption moved from dedicated hardware into software. The program outlived its exposure; the documents do not say when or why it ended. According to the Post, the headquarters building was sold in 2017, and in 2018 what remained of the company was split and sold through the same Liechtenstein law firm that had arranged the purchase in 1970, in transactions the paper said seemed designed to provide cover for a CIA exit; current and former officials estimated the 2018 sales at between 50 and 70 million dollars.
Three Records
A secret like this one exists in several records at once, and they are not equal.
The first record was the press. By 1996 it contained almost everything a careful reader needed, except an account from someone with the authority to confirm it, and so it remained an allegation the company could deny and its customers could choose to disbelieve.
The second record belonged to the operators. In 2004 the CIA's Center for the Study of Intelligence completed a classified internal history of Minerva, just under a hundred pages long. In 2008 the BND compiled an oral history of its own role. Written by and for the agencies, they contained what the 1995 article could not: who owned the company, for how much, what the machines delivered, and how the customers behaved when the story nearly broke. The Swiss parliamentary inquiry later found that the German side had apparently received a copy of the CIA history around 2005, and that this version is the one that eventually reached the press.
Even that record is partial. The Swiss delegation found that the CIA history ends in the mid-1990s and that its passages on Switzerland are often imprecise and in places demonstrably wrong. And it never fully changed hands. When the Post, ZDF and SRF reported on it in February 2020, they quoted selected passages, because, the Post wrote, its source insisted that only excerpts be published, and in November that year the Swiss delegation noted that the history itself had still not been made public. The CIA and the BND declined to comment, and American and German officials did not dispute the documents' authenticity. What reached the public came by leak, and in excerpts.
The third record belonged to Switzerland, and it is the one that most clearly earns the name of this series.
Two Offices in Bern
On 24 March 1994, the day after Bühler's television appearance, the head of the Swiss federal police, Urs von Daeniken, assured the parliament's Control Delegation, known by its German initials GPDel, that his service would establish who the real owners of Crypto AG were. The inquiry had two questions: who controlled the company, and whether its machines had been manipulated. Between March and November the police questioned Bühler a second time and twenty current and former employees and board members. Two of them described alleged manipulation.
The investigators got close. When two members of the company's advisory board gave unsatisfying answers, the justice ministry's secretary general wrote on 5 May 1994 to Georg Stucky, a member of parliament who had sat on Crypto AG's board since 1992, urging him to use his influence to clarify the ownership. According to the CIA history and other files the delegation later obtained, the company's management had told Stucky that spring who the true owners were; Stucky told the delegation in 2020 that he had not known he worked for a firm owned by American intelligence. The cooperation the ministry sought, the delegation wrote, evidently never came about. On 27 May 1994 the police chief reported to the overseers that there was no concrete suspicion of a crime, but that the police had indications possibly pointing to two quality levels in the encryption, and that the trail of ownership ended at "a Liechtenstein foundation with West German capital participation." On 3 May 1995 the police wrote in their final report that the true owners of Crypto AG could not be identified.
The answer was already in Bern. According to the delegation's report, the Strategic Intelligence Service had obtained reliable information from the autumn of 1993 that the company was owned by foreign intelligence services. It set out to break the weakened machines itself and, years later, with American consent, obtained the technical knowledge to do so. Inside the service the knowledge was, in the delegation's words, a well-guarded secret, held by its head and one or two others. The overseers who were briefed on the police inquiry heard nothing of what the intelligence service knew.
The delegation found something more specific than silence. Once the police inquiry had begun, it wrote, the intelligence service should have handed over what it knew. Instead it told the police that it had no indication that foreign intelligence services stood behind Crypto AG. The delegation judged that the service had placed its own collection and its relationship with the Americans above the interests of the prosecution, a choice it had no authority to make; the weighing, it said, belonged at the political level. It found the police inquiry correctly conducted on the information available to it, and no evidence that politicians interfered with it. One office of the state asked the question, another held the answer and said it had none, and the body charged with overseeing both was told about the first and not the second.
The technical warning was older still. In 1983 the federal signals office, after examining a Crypto machine, concluded that the allegations of manipulation could be neither proven nor definitively refuted, and noted a discrepancy between the machine's logic and its documentation. The files do not show how the police and prosecutors responded; the signals office turned its attention to the security of the machines Switzerland itself used.
The delegation published its findings on 10 November 2020. It found the intelligence service's cooperation in principle lawful, and its political significance very high. The defence minister, Viola Amherd, was briefed by the intelligence service in August 2019, and the Federal Council, Switzerland's government, in November 2019, three months before the story became public. None of Amherd's predecessors had been told by the intelligence service. The CIA history claims that one of them heard about the owners from a Crypto board member during the Bühler affair; that former minister, Kaspar Villiger, told the delegation he believed he had spoken with that board member about Crypto AG during the affair, but had no memory of ever being told who owned it. An intelligence service note from March 1994, the delegation found, shows that he was kept in the dark about the owners while the service's chief knew.
The old service's files had survived in a converted underground facility. The delegation also found that between 2011 and 2014 the current Federal Intelligence Service had destroyed records of the former domestic service's exchanges with foreign partner services; the report does not establish that these concerned Crypto AG. It concluded that the Swiss authorities shared responsibility for Crypto AG's activities, and that because the intelligence service's leaders kept the information to themselves, successive defence ministers were never in a position to exercise theirs.
Then the record closed again. The Federal Council had first commissioned a former federal judge, Niklaus Oberholzer, to investigate; the delegation then took over his inquiry, and his report, about ninety pages long, was written for the delegation and the Federal Council only and classified because, the delegation said, its contents could seriously harm the country's interests. When access was later sought to the earlier report he had written for the Federal Council, the federal transparency commissioner in August 2022 backed the refusal, on the ground that the document belonged to the Federal Council and fell outside the transparency law. As far as I could establish, neither report has been published.
What followed the revelations landed mostly on the companies that had inherited the name. The economics ministry suspended the general export licences of the two successor firms in December 2019, a move the delegation later called unjustified in substance and in law. A criminal complaint filed in February 2020 was dropped that December, and about 400 seized devices were returned. In June 2020 Crypto International, which had bought the international business and the brand in 2018, said it had to dismiss nearly all of its staff in Switzerland, 83 people. In May 2021 the head of the Federal Intelligence Service, Jean-Philippe Gaudin, agreed with the government to leave.
Read in sequence, the Swiss record has a shape. The state knew in 1993, and in 1995 another office of the same state wrote that it could not find out, after the office that knew had told it there was no indication. Its political leadership learned in 2019. The fullest account the state later commissioned is still closed. The penalties fell on people who had bought a brand in 2018 and learned in 2020 what it had been, among them the 83 employees who lost their jobs.
The Record Has to Change Hands
In February 2020 a record changed hands, and only in part.
Much of the 2020 reporting would not have surprised a reader of the Baltimore Sun in December 1995. The arrest, the bail, the rigged algorithms, the NSA consultant at the design meeting: all of it had been in print for a quarter of a century. What was new was proof of the ownership, and the source that supplied it. The story now came with the operators' own account, in their own words. After 2020, the company's old denial would have had to contradict the agencies' own history. The Sun had said in 1995 that this was what it would take.
Look at who did not know. The salesman, most of the engineers, the customers' own cryptographers, the Swiss police, successive defence ministers: each was missing a different part of the same fact, and each could therefore deny, investigate or govern in good faith.
That, in my reading, is the mechanism this case shows. Public reporting raises the cost of a covert arrangement, but it settles the question only when the authoritative record leaves the people who hold it. Until then, public knowledge has to compete with a denial, and for customers who have no easy alternative the denial is enough to keep buying, or at least enough not to stop. In this case exposure did raise the cost: six or more customers left, and Germany, rattled by the Hydra affair, sold out. The distance between an allegation and an authoritative record was one support among several; inertia and American money did much of the work. What the distance did was make survival easier, and the distance was built and maintained: through bearer shares, a single office that decided what the algorithms were, a lawsuit drawn out until it could be settled, and a chief executive willing to deny what he knew.
A secret is not over when it is published. It is over when the record changes hands.
Even then, a record that changes hands has to be tested before it settles anything. Three stages are easy to confuse: publication, a claim appearing in print; disclosure, the record leaving those who hold it; and verification, someone independent being able to test it. Rubicon reached the first in 1994 and 1995 and the second only partly in 2020. It has not reached the third, since the CIA history surfaced in excerpts and the Swiss inquiry found it wrong in places. What changing hands does is end the monopoly on the record, and move the argument from whether the documents exist to what they show.
The point carries well beyond Zug; in November 2020 another Swiss firm, Omnisec, was alleged to have sold weakened devices. Classified histories, internal audits and commissioned inquiries are written by the institutions that ran the programs they describe, for readers inside those institutions. They are often the only documents complete enough to settle what happened, and that is why they stay inside. The CIA history of Minerva left the agencies' hands in excerpts. Oberholzer's reports have not left at all.
The Strongest Objection
The strongest counterargument to this reading comes from the operators' documents themselves. According to the Post, the intelligence kept coming after 1995 partly because of bureaucratic inertia: many governments, particularly in less developed countries, never got around to replacing their machines. The company stayed in business only because the American government kept paying for it. Inertia explains why the source survived; subsidy explains why the firm did. Remove the gap between allegation and record, and neither changes. Iran, the case this piece leans on, may have bought again for reasons that had nothing to do with the machines at all, if the CIA was right that Bühler was taken as leverage. When the documents leaked in 2020, the security technologist Bruce Schneier wrote that it "isn't really news," and the leak settled strikingly little: the company had already been sold off, the Swiss criminal case was dropped, and the most detailed Swiss report stayed secret. The Swiss evidence is also the weakest support for design at the level of the state: the delegation judged the police inquiry correctly conducted on the information it had, found no political interference, and traced the silence to a handful of people inside one service. On this view the thesis restates what the Sun's reporters already knew in 1995.
Much of this is right, and it explains the cancellations, the survival of the business and the end. The reading offered here does not claim that the leak of 2020 ended the operation, whose company had been sold off two years earlier, or that the gap between allegation and record was the main reason customers stayed. It claims something narrower: that the gap was held open by design, and that holding it open was worth the effort to the people who ran the program. The objection explains why individual customers stayed. It does not explain why a lawsuit was drawn out and then settled on terms that silenced the man it had been brought against, why a police inquiry in Bern could not find what an intelligence service in Bern already knew and told it otherwise, or why, in Switzerland, the fullest account remains closed today. A decision taken by a handful of people inside one service is still a decision, and the delegation found that the service had no authority to take it; the police, meanwhile, had been working on less information than the state held. Both readings can be true at different levels: customers may well have stayed for ordinary reasons, while the record stayed inside for structural ones.
The Files
Hans Bühler came home on 5 January 1993 believing that his employer had been falsely accused and that he had paid for it with nine and a half months of his life. Within a year he had begun to suspect otherwise. He said so on television in March 1994 and to a reporter from Baltimore in September 1995, and the settlement that ended the company's lawsuit required him to stop. He died on 27 July 2018, the year what remained of the company was split and sold, a year and a half before the documents appeared that would have told him what he had been selling.
In early November 2019, the Federal Intelligence Service made further copies of its internal assessment of the Crypto affair, the document that explained who had owned the company and what the Swiss service had known. One went to its acting head of collection. He was, the delegation noted, the man who had led the federal police inquiry in 1994. A quarter of a century after his investigators reported that the owners could not be identified, he was handed the answer by the successor of the service that had told his inquiry it had no indication.
In the end, everyone in this story had a file: the CIA its history, the BND its oral history, the Swiss intelligence service its records in a converted underground facility, the Swiss police a final report that could not name the owners, and the Federal Council a report it has not published.
The customers had the machines. And the man who sold them had a settlement that required him to declare there had been nothing to find.
Evidence Map
Facts, interpretations, forecasts, and disconfirming signals.
Core claim. Operation Rubicon outlived public exposure in 1994 to 1996 while the authoritative record stayed with the agencies that ran it and, in Switzerland, inside one office of the state, which told the police it had no indication of foreign ownership; that gap was built and maintained by design. The operation's survival also rested on customer inertia and American subsidy, and the decisive documents reached the public in 2020 by leak, in excerpts, not by release.
Evidence level. Facts: high for the ownership, the Bühler case, the 1995 and 1996 reports and the Swiss findings (CIA history as reported by three newsrooms and not disputed by American and German officials; GPDel report; contemporaneous press). The lawsuit strategy and settlement terms rest on the Crypto Museum's account of the CIA history, corroborated in outline by the Sun and Der Spiegel. Interpretation: medium for the mechanism that exposure ends an arrangement only when the authoritative record changes hands. Single-source figures from the CIA history (the 40 percent share, Iran's reading rate, the cancellations) are marked as such in the text.
What would confirm this. Release of further operator records (the full CIA history, the BND oral history, Oberholzer's reports) showing that officials weighed disclosure and chose not to. On the Swiss side this is already partly documented: the GPDel found that the intelligence service did not hand its knowledge to the police inquiry and instead told the police it had no indication of foreign ownership.
What would disprove this. Customer sales data showing a sharp, report-driven collapse after 1995, so that exposure alone was effective and the program survived only on subsidy; evidence that the intelligence service's knowledge did reach the police and was set aside there; publication of Oberholzer's reports, which would remove the Swiss part of the claim that the record is still closed.
Watchlist. Any Swiss decision on access to Oberholzer's reports; any official CIA or BND acknowledgment or release of the Minerva history; litigation by the successor companies; similar cases of state-linked suppliers (the Swiss firm Omnisec was alleged in November 2020 to have sold weakened devices).
Sources. The Washington Post investigation of 11 February 2020 (with ZDF and SRF), the GPDel report on the Crypto affair, the Baltimore Sun of 10 December 1995, Der Spiegel of 1 September 1996, Res Strehle's Verschlüsselt (1994), and the Crypto Museum's files on Rubicon and Hans Bühler.
Jerry van der Laan writes The Manifest Archive, Forensic Narrative Intelligence Writing on the systems beneath power, money, and history. He traces the structures beneath them.
