By December 2010, Uganda was supposed to have three and a half million national identity cards. It had 401. One belonged to the President. One belonged to the Prime Minister. The equipment bought to make the rest sat at the state printing corporation in Entebbe, uninsured, and before long some of it was stolen or damaged.
Parliament investigated and found that officials had flouted procurement law in hiring Mühlbauer that March, and that the cost had been inflated from about Ush200 billion to Ush249.5 billion without explanation. There had been no open bidding. The contract was reportedly issued on President Museveni’s instruction after he met the German ambassador at State House.
Then the system worked. Mass enrolment began in 2014, and within five years close to eighty percent of Ugandan adults were carrying national identity cards. The World Bank’s own country diagnostic called it the fastest identity rollout on the continent, and Mühlbauer put out a press release calling it a world record.
A system that failed outright would have been abandoned. This one recovered and was declared a success, so it was kept, and everything that came after had to be argued against that verdict.
A government that cannot leave its vendor cannot govern its infrastructure. It can only inhabit it.
Uganda never owned what it had bought. When Parliament asked NIRA’s executive director why the country still could not issue identity cards to legal foreign residents, she explained that the alien register had never been built into the system, which was designed for citizens and adult registration and nothing else. And in 2019, when the first contract expired, Uganda bought a new system and still did not obtain the intellectual property rights.
Two contracts. A decade. A state that could not add a category of person to its own register of persons.
In 2022 the National Identification and Registration Authority told Parliament that the card printing equipment Mühlbauer had left behind was inflexible and vendor-locked, and that this was blocking the upgrade the agency needed. Mühlbauer’s vice president publicly denied it. Parliament opened a probe, and one member of the committee said they were being fed lies by one side or the other and had no way to tell which.
The lock-in may or may not have been real, but what is certain is that twelve years and roughly a hundred million dollars into its national identity system, the Ugandan state could not establish the truth. Answering that question required the capacity to read its own infrastructure, and Uganda had contracted that capacity away in 2010, in a deal nobody was allowed to bid against.
Okye was 88 when researchers found him in Namayingo, in eastern Uganda. When he was registered, there was an error in his date of birth, and the register made him 79. Uganda pays a cash transfer to citizens over 80, which Okye was, and which the register said he was not. Senior sources told the researchers documenting his case that at least fifty thousand Ugandans over 80 are dealing with the same kind of error, or have no card at all.
No vendor did this to Okye. A field was entered wrong and became impossible to correct, because fixing it at scale meant changing a system the state could not change.
India, on the other hand, saw this coming. Aadhaar, India’s national identity system, was built in-house, with public infrastructure as the operating assumption, and the Indian state owns it in a way almost no equivalent is owned. When other governments tried to copy it they found they could not, because Aadhaar was never a product, so the team behind it built MOSIP, an open-source version, and gave it away. Morocco, the Philippines and Ethiopia have all adopted it. The pitch was clean: if you are going to enrol your population, the code should belong to you.
It should, but even that is barely enough. During the 2019 update of the National Register of Citizens in Assam, an exercise to determine who in the state is legally Indian, the Indian government locked the biometrics of 2.7 million people who had filed claims, as a matter of procedure. No biometrics meant no working Aadhaar, and no working Aadhaar meant no rations, no pension, no bank account. When unlocking began five years later, the government released 900,000 people and left 1.8 million locked, with no explanation of what separated them.
Sovereignty over the code is necessary. It does not, by itself, produce sovereignty over the consequences.
India owns every line of the Aadhaar code. That bought the ability to build, and it did not buy the ability to correct.
The procurement document for any major identity project should answer four questions, and almost none of them do. What can the state change without the vendor’s consent. What can it see inside the system. What happens if the vendor fails or falls under another government’s control. And what would migration cost, in money and in years.
These questions go unasked because asking them honestly makes the project harder to approve. Honest answers produce a longer timeline, a higher cost, and a contingency plan nobody wants to fund. Ministers need a launch date and finance teams need a savings narrative, and the vendors have learned to answer the question that was asked instead of the one that should have been.
Even when asked and answered, the four questions only cover the moment of purchase. The system will be in production for decades. The population it describes will change, categories of resident will appear that nobody anticipated, and errors will accumulate in the records of people who have no way to reach the people who could fix them. Procurement is a decision taken once. The infrastructure is a decision taken continuously, by whoever has the standing to change it.
This is the capacity most states never build, and it is the one that governs everything downstream. Digital infrastructure moves faster than the policy that authorises it, which means the policy framework has to be as changeable as the system it governs, or the system stops answering to anyone. A government that cannot amend its own register is not administering it. It is hosting it on behalf of whoever can.
The people in the register are the owners. The government works for them, and the infrastructure is the embodiment of that work, which makes the ability to change it the whole substance of the arrangement.
The debate about digital public infrastructure is conducted in the language of ownership: source code, data residency, national control. In 2020, Okye needed something smaller than that. He needed a state that could correct a number in a field while he was still alive to collect.