> ## Content Index
> Fetch the complete content index at: https://www.ceo.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# The machine and the system: IBM, apartheid, and the question of corporate complicity
- URL: https://www.ceo.com/blog/the-machine-and-the-system-ibm-apartheid-and-the-question-of-corporate-complicity/
- Published: 2026-03-30T03:15:00.000Z
- Updated: 2026-08-19T03:18:35.000Z
- Description: IBM maintained that it could not control how its products were used.
- Author: CEO.com
- Tags: #blog

By 1952, IBM was selling electronic tabulators to the apartheid government of South Africa. By the 1960s, it was actively bidding on contracts to build population registration systems that would track millions of people by race. By 1982, it acknowledged in a letter to the U.S. State Department that its computers were being used for South Africa's national identity system. By 1987, it had earned hundreds of millions of dollars from a government the international community had formally sanctioned for crimes against humanity.

Throughout it all, IBM maintained that it could not control how its products were used.

That position ultimately prevailed in court. In 2015, the Second Circuit Court of Appeals dismissed the claims of apartheid survivors who had sued IBM under the Alien Tort Statute, finding insufficient evidence that IBM had acted with the purpose of facilitating apartheid. In 2016, the Supreme Court declined to hear the case. IBM was legally free.

What IBM was not, and has never been, is historically free. The case study it left behind is the most instructive precedent in existence for technology companies navigating the question of what they owe when a powerful government wants to use their tools in ways that harm people. That question has never been more urgent than it is today.

### **The Pattern Before the Pattern**

IBM's relationship with the apartheid government did not emerge in a vacuum. In the 1930s and early 1940s, IBM provided Hollerith punch card systems to its German subsidiary for use by the Third Reich. Historian Edwin Black documented the arrangement exhaustively in [IBM and the Holocaust](https://www.amazon.com/IBM-Holocaust-Strategic-Alliance-Nazi/dp/0914153277?ref=ceo.com). The systems were custom-designed to track and sort groups targeted for persecution. The identification numbers tattooed on Auschwitz inmates began as IBM punch-card system numbers. IBM president Thomas J. Watson Sr. received the Merit Cross of the German Eagle from the Nazi government. When U.S. law prohibited dealing with Germany after the war began, IBM chose concealment over withdrawal.

The systems IBM later built in South Africa shared the same structural DNA. In both cases, racial classification was the primary identifying characteristic. In both cases, the equipment was leased rather than sold, meaning IBM always knew which agencies were using the machines. In both cases, the technology required close, customized collaboration between IBM and the government. In both cases, IBM said it was not responsible for what governments did with its products.

This was not a company that stumbled into complicity once and failed to recognize it. When confronted with a government seeking to identify, classify, and control a population by race, IBM made itself available. It had done it before. It did it again.

### **The Architecture**

Apartheid required administration on an industrial scale. The South African state maintained two parallel population registration systems. The Department of the Interior managed the [Book of Life](https://scielo.org.za/scielo.php?script=sci%5Farttext&pid=S0259-01902014000100010&ref=ceo.com), a comprehensive identity document covering citizens classified as non-black. That system ran on IBM hardware. The Plural Affairs Department maintained a separate passbook system tracking more than 25 million Africans classified as black, recording racial classification, name, sex, date of birth, residence, photograph, fingerprints, marital status, and every movement in and out of the country.

IBM had competed in 1965 for the passbook contract covering the Black population and lost. It won the Book of Life contract instead. The company was not a reluctant supplier caught up in something it had not sought. It was an active bidder for apartheid contracts. It won one and pursued others.

In 1970, the South African government relaunched its population register on new IBM hardware, constructing a 30-story building in Pretoria called Civitas specifically to house it, with a dedicated IBM computer room in the basement. By 1982, the South African police were making 850,000 annual queries at display screens connected to the population registry, examining nearly 180,000 fingerprints and making more than 13,000 positive identifications. IBM equipment was present in the Department of Labor, the Treasury, the Reserve Bank, and the Department of Prisons. The South African Parliament passed legislation making it illegal for companies to report the use of their computers to outside parties.

As one [analysis of the period](http://www-cs-students.stanford.edu/~cale/cs201/apartheid.comp.html?ref=ceo.com) concluded, "Computer technology did not merely support the system of oppression, rather the entire country was utterly dependent upon it."

### **The Defense**

IBM's position throughout was consistent: it could not control how customers used its products. Annual South African sales reached approximately $300 million by 1982 and an estimated $450 million by 1986, much of it going directly to the South African government. IBM was the largest computer supplier in South Africa throughout the apartheid era. It signed the [Sullivan Principles](https://philadelphiaencyclopedia.org/essays/sullivan-principles/?ref=ceo.com) in 1977, a voluntary corporate code of conduct that committed American companies in South Africa to equal pay and integrated workplaces. IBM paid its Black employees equally. It invested in training and education programs.

None of that had anything to do with what the technology was being used for. The Sullivan Principles addressed behavior inside the corporate workplace. They said nothing about what governments were permitted to do with a company's products once those products left the building. IBM met its workplace commitments and simultaneously provided the systems that helped the state classify, track, and control the people those employees went home to at night.

IBM formally divested from South Africa in 1987, the same year Leon Sullivan abandoned his own principles, declaring that a decade of constructive engagement had accomplished nothing and calling for full corporate withdrawal. IBM's divestment was structured as a management buyout. The subsidiary continued to operate. IBM products remained. Licensing revenue continued. The machines stayed.

### **What the Courts Said**

In 2002, survivors of apartheid-era violence, represented by the [Khulumani Support Group](https://www.business-humanrights.org/en/latest-news/apartheid-reparations-lawsuits-re-so-africa/?ref=ceo.com), filed suit against 23 corporations that had done business in South Africa during the apartheid era. IBM was among them. The plaintiffs included relatives of people tortured, raped, and killed in connection with passbook violations, violations made possible in part by the identification systems IBM's technology had helped build.

The litigation lasted 14 years. In 2015, the [Second Circuit Court of Appeals](https://foleyhoag.com/news-and-insights/blogs/global-business-and-human-rights/2015/july/alien-tort-case-development-second-circuit-affirms-dismissal-of-claims-against-ford-and-ibm/?ref=ceo.com) affirmed the dismissal of claims against IBM. The court's reasoning is the part every technology executive should read. The court acknowledged that IBM had acted with knowledge that its technology might facilitate apartheid. It found that knowledge was not enough. To establish liability, the plaintiffs needed to show purposeful facilitation. The court held they had not. "Plaintiffs do not and cannot plausibly allege that by developing hardware and software to collect innocuous population data, IBM's purpose was to denationalize black South Africans and further the aims of a brutal regime."

The Supreme Court declined to hear the appeal in 2016\. IBM was legally clear.

The court had drawn the line on purpose. IBM knew. IBM did not intend. Under the law, that distinction was sufficient. Under any other framework, it is harder to sustain.

### **The Question That Remains**

The subsidiary structure provided legal cover. IBM South Africa had signed the contracts. IBM South Africa had installed the machines. IBM South Africa had trained the operators. IBM New York had set a strategy, provided technology, collected revenue, and maintained the relationship across decades. The parent and the subsidiary were legally distinct. That distinction was large enough to defeat a lawsuit. It was not large enough to change what happened to the people the system was used against.

IBM did not set out to administer apartheid. Its machines worked exactly as designed. Population data was collected. People were classified by race. The system identified them. The state controlled, arrested, and in many cases killed them. The technology was not the weapon. It was what made the weapon precise.

The "we cannot control how our products are used" defense is now the default posture of the technology industry. It is invoked by social media companies when their platforms are used to incite violence. It is invoked by surveillance technology companies when their tools are used by authoritarian governments. It will be invoked by AI companies when their models are used in ways that harm people. IBM did not invent this defense. It perfected it over two decades and two continents and prevailed in court with it.

But here is what the IBM case actually demonstrates: the defense works legally precisely because it is built on a distinction between knowledge and purpose that is morally incoherent. IBM knew what its technology was being used for. It knew who was using it. It had leased the machines, not sold them, which meant it maintained continuous relationships with the agencies operating them. The distance between knowing and intending, in that specific context, is not a distance any serious person should find meaningful.

That distinction is where today's AI industry lives. The companies building the most powerful AI systems in history know, or can know, what their technology is capable of. They know which governments are buying it and what those governments have done with technology before. The question of whether they intended the specific harm that resulted is one that courts may ultimately find in their favor. It is not a question that history will.

Anthropic drew a line. It said its technology would not be used for mass domestic surveillance or fully autonomous weapons. The government called that a supply chain risk. A federal judge called the government's response illegal retaliation. The argument about where the line belongs is genuinely difficult. The argument about whether there should be a line is not. IBM spent 35 years proving what happens to companies that decide there should not be one.

The machines worked. The system functioned. Millions of people were classified, tracked, controlled, and harmed. IBM collected the revenue and said it was not responsible. The courts agreed.

That is one way to run a technology company. It is not the only way. And it is not, whatever the legal outcome, a defense that holds.