On 11 February 2020, the Washington Post and Germany’s ZDF published a joint investigation that rewrote a chunk of twentieth-century intelligence history; Swiss broadcaster SRF aired its side of the story a day later. An encryption company based in Zug, which had sold its machines to more than 120 countries since the 1950s, secretly belonged to the CIA and West Germany’s foreign intelligence service, the BND. For nearly half a century, dozens of states encrypted their diplomatic and military communications with equipment two Western agencies held the key to.

That company, Crypto AG, really did exist this way, and its history is now acknowledged by the very governments that ran it. It raises a question few threat models spell out: a foreign intelligence service can place or turn people into management, security or research roles without ever breaching a firewall. Is that risk, in your organisation, accepted, mitigated, or simply forgotten?

In shortInsider risk, a danger that comes from inside the organisation rather than an external attacker, goes well beyond the disgruntled employee copying files on their way out. Several well-established cases prove it: Crypto AG (a fact acknowledged by the governments involved), fake North Korean IT workers (a highly active current threat), and harder-to-settle suspicions like Gemplus, not to be presented as fact. This risk belongs in the threat model alongside phishing and ransomware, with hiring screens, access separation, and reporting doubts without shame as the best defenses.

Crypto AG: the coup of the century

Boris Hagelin set up Crypto AG in Zug in 1952. An informal deal with the NSA existed as early as 1951 and was formalised in 1955. In June 1970, the CIA and the BND secretly bought the company for $5.75 million; the BND would sell its stake to the CIA in 1993, for $17 million.

The mechanism was nothing like a crude backdoor. The NSA wrote alternative operating manuals that led customers, without their knowledge, to configure their own devices with reduced cryptographic complexity, and designed the algorithms of the more advanced models, like the H-460, directly. Client countries trusted Switzerland’s image of neutrality, never suspecting who actually sat behind the ownership structure. The consequences reached official speeches: in 1986, after the La Belle discotheque bombing in Berlin, Ronald Reagan publicly cited intercepts as evidence against the Libyan regime.

Hans Bühler, Crypto AG’s chief salesman, was arrested in Tehran in March 1992 and held for nine months, suspected of espionage. The company paid roughly a million dollars for his release, then fired him shortly after his return. The story broke wide open in 2020, based on a classified 96-page internal CIA history that calls the operation “the intelligence coup of the century”. On 10 November 2020, the Swiss Parliament published a report establishing political co-responsibility on the part of Swiss authorities, who had known since 1993.

This is a long way from the usual script of an infiltrated agent climbing an org chart. The company itself was captured, through ownership of the capital and control over hiring for key positions. Only a handful of executives and engineers were in on it; the rank and file, Bühler included, had no idea what they were part of.

A risk that hasn’t gone away

Phishing and ransomware sit in the threat model of virtually every company that has one; a foreign intelligence service placing or recruiting someone into a strategic role rarely makes the list. Yet it’s a patient, well-funded operation. Access comes legitimately, with a badge and a signed contract, and the target is a position rather than a server: the one that holds the encryption keys or the product roadmap. The most exposed sectors (cryptography, semiconductors, industrial automation, IoT, energy, defense, telecoms) put Switzerland high on the list, because of the neutral reputation Crypto AG traded on for decades.

The Gemplus suspicion

The Gemplus case sits at a very different level of evidence than Crypto AG. Nothing here is established: suspicions raised at the highest political level, never examined or proven. Co-founded in 1988 near Marseille by Marc Lassus, the company became the world’s leading smart card manufacturer within a decade. In early 2000, the American investment fund Texas Pacific Group (TPG) took a 26% stake. Pushed out of the chairmanship in late 2001, Lassus resigned from the board in December 2002, just before the shareholder meeting convened to remove him. A few months earlier, Alex Mandl, a former AT&T president and COO, had taken over as CEO, leaving the board of In-Q-Tel, the CIA’s venture capital fund, as he did so. A resolution filed in the French Senate in 2003, describing those In-Q-Tel ties as “established”, was never examined.

No source documents a proven transfer of encryption keys or data to a foreign service to this day. What is confirmed is ordinary governance: a stake in the capital, weight on the board, a well-connected executive, the founder eased out by the book. Nothing illegal, at any stage. And that’s the problem: no antivirus will ever flag a takeover through capital.

North Korea’s fake IT workers

Since at least 2022, the North Korean regime has been sending workers posing as candidates of other nationalities to land remote IT jobs at Western companies. The U.S. State Department, Treasury and FBI documented the practice as early as May 2022 in a joint advisory, already flagging the sums flowing back to the regime, down to its weapons programs. The U.S. Department of Justice now counts more than a hundred companies affected, including Fortune 500 groups. Those diverted salaries come on top of the regime’s cryptocurrency thefts: around $2 billion in 2025 according to Elliptic and then CrowdStrike, including $1.46 billion from the Bybit exchange hack alone. That one was pure hacking, not the IT worker scheme.

The mechanism is documented across numerous court cases: stolen or fabricated identities, AI-generated résumés and, increasingly, video interviews faked in real time with deepfakes. In July 2026, the FBI and the State Department signed an alert with ten partner countries that devotes a full section to those faked interviews. Once hired, the worker has their laptop shipped to a US-based facilitator, who hosts the machine and installs remote access on it. The real operator then works “from” a US IP address. Sometimes the arrangement escalates to extortion: stealing the source code, then threatening to leak it if payments stop.

Christina Chapman, in Arizona, was sentenced in July 2025 to 102 months in prison for running such a “laptop farm” between 2020 and 2023: $17 million generated, more than 300 companies affected. In July 2024, the security company KnowBe4 disclosed that it had mistakenly hired a fake North Korean engineer despite four video interviews; access was cut 25 minutes after the laptop arrived, with no data compromised.

The warning signs catalogued by the FBI and its IC3 center apply to any remote hire: reluctance to turn on the camera, near-identical résumés across candidates, frequent changes to the shipping address or the bank account for salary payments, VPN connections inconsistent with the stated location. Then there is the hand-wave test on video. Real-time deepfake software struggles to follow a hand passing quickly in front of the face, and the image visibly glitches. The most recent tools are gradually closing that gap, though.

Recruitment or placement

In counterintelligence jargon, recruitment and placement describe two distinct operations. Recruitment means turning someone already in post, drawing on one of the levers practitioners sum up with the acronym MICE: Money, Ideology, Coercion, or Ego. Placement inserts the agent upstream, on the model long used by Russian services for their “illegals”; the FBI dismantled ten of them in one sweep in June 2010, Anna Chapman among them. These networks almost always target political, diplomatic or military circles; when their agents work in business, finance or real estate, it’s mostly as cover, rarely with the company itself as the end target. For most businesses, the likelier scenario remains the recruitment of someone already in post.

What a company can actually do

The UK’s NCSC notes a widening gap between the pace of the threat, criminal and state-backed alike, and organisations’ resilience. A gap it traces first to governance: cyber risk stays delegated to technical teams instead of being owned by the board. As Nicole Perlroth, the former New York Times cybersecurity reporter, put it in 2025: “Ransomware hits the bottom line this quarter. [State espionage] hits it five years from now.” In Switzerland, a University of Bern study for the FIS, published in 2020 after surveying 362 companies, found that around 15% of firms describe themselves as targeted by economic espionage (up to 35% in the subsample interviewed in depth). For 11% of affected companies, the incident threatened the firm’s very existence. The EU’s NIS2 directive now makes boards accountable for security measures, down to human resources security and access control: insider risk becomes an auditable item in practice. It still draws no distinction between the nation-state agent and generic insider threat.

Hiring screens are the first line, but they’re never enough on their own. For any remote hire, the FBI recommends scrutinising identity documents, through a verification service or an in-person check where needed, and contacting past employers directly rather than relying on references alone. The video interview, camera on, should include the test described above: have the candidate wave a hand in front of their face.

That initial screening has a limit, though, established by the UK’s NPSA (the former CPNI): 76% of recorded insider incidents were committed by employees with no malicious intent at the time they were hired. Their loyalty shifted afterward, under the weight of debt, a rough termination, or coercion. Hence the case for continuing the vetting over time rather than betting everything on the job interview.

The rest is about access architecture more than hiring. Least privilege and separation of duties limit what any single role, however senior, can do alone; the four-eyes principle (no critical action without sign-off from a second person) applies to irreversible operations: recovering a master key, changing admin rights, accessing a production environment.

Behavioural monitoring deserves a word of caution, particularly in Switzerland. Article 26 of Ordinance 3 to the Employment Act, together with the nLPD (and the GDPR on the European side), strictly regulates this kind of control: exceptional, proportionate, disclosed to employees, never used to evaluate performance. What remains is the human dimension of weak signals. The CPIR model (“Critical Pathway to Insider Risk”), published in the CIA’s Studies in Intelligence, maps the typical path to the act: personal predispositions (personality traits, a history of rule-breaking), stressors such as debt, a workplace conflict or a termination, then concerning behaviours colleagues can notice. The right organisational response can break that progression in time; a clumsy one speeds it up. A single isolated signal proves nothing. Several signals recurring over time, on the other hand, are worth flagging to the security team, following the same principle already described for reporting a cyber incident.

For an SME or a young Swiss company, all of this fits into an affordable core: a systematic video interview, third-party identity verification for a few dozen francs, siloed access from day one. Switzerland’s Federal Intelligence Service has run a prevention programme in this spirit since 2004, Prophylax: confidential bilateral interviews with exposed companies and universities, paired with an awareness film, En ligne de mire. Its director nevertheless resigned in 2025, castigating the service’s lack of resources and staff in his departure letter; Parliament had just swept aside his request for 150 additional posts.

Keeping perspective

The vast majority of international hires are exactly what they appear to be. No one gains from treating every new board member or foreign candidate with suspicion on the strength of their passport alone. The examples in this article point to Washington, Berlin, Moscow or Pyongyang because those cases are documented through court records, parliamentary reports or press investigations; that says nothing about those who haven’t been caught. Any state with an intelligence service can take advantage of an insider, declared adversary or long-standing partner.

Crypto AG and the current wave of fake North Korean workers mostly show that an intelligence service has the time and the budget to play a decades-long game, without the pressure of a ransomware crew that needs to cash out fast. Crypto AG took nearly fifty years to come to light; the laptop farms had been running for years before most of the companies involved heard of the scheme. Better to spot that kind of setup while a process fix still covers it than to wait for a prosecutor to do it for you.

Sources