Korea moved almost all of government online and found the limits of doing so

Korea’s public administration is, by the standard international measures, among the most thoroughly digitised anywhere. The United Nations E-Government Development Index has placed the country third in the world in successive editions, and the government reports that the great majority of civil services — residence certificates, family relation records, tax documents, business registrations, benefit applications — can be completed through the Government24 portal without speaking to anyone. For most ordinary transactions this is simply true, and the experience is fast in a way that surprises visitors from countries with comparable incomes.

The foundations were laid earlier than the interface suggests. Korea built out broadband in the late 1990s, ran a series of national informatisation programmes through the 2000s, and — decisively — had already assigned every resident a single permanent identification number. Joined-up service delivery, the thing most e-government projects fail at, is straightforward when every database already shares a primary key. That design choice explains both the convenience and the country’s unusually severe exposure to data breaches, since the key cannot be rotated once leaked.

The convenience layer kept extending. Document issuance moved to unattended kiosks, then to phones. A mobile driver’s licence launched in 2022. The MyData regime, effective in finance from January 2022, obliged institutions to hand a customer’s records to a competitor on request, importing the portability logic of public records into the private sector. A notification service pushes reminders about expiring documents and eligible benefits. Each addition made the digital path more attractive and the physical path more vestigial.

Then, in November 2023, the administrative network used by local governments failed. Government24 went down, and it stayed degraded for days. The eventual diagnosis pointed to faulty network equipment rather than an attack, which was reassuring about intent and not at all reassuring about resilience. The visible consequence was that district offices could not issue documents at the counter either: the counter had long since become a person operating the same system a citizen would have used at home. The offline fallback had not been withdrawn by policy. It had simply atrophied from disuse, and nobody had rehearsed it.

The structural explanation is procurement as much as engineering. Korea’s public systems accumulated over three decades as hundreds of separately commissioned projects, built by different contractors, integrated at the national data centres. Fixed-price competitive tendering rewards the lowest bid; a rule adopted in 2013 excluded large conglomerates from most public software work in order to protect smaller firms, with the side effect that very large integration projects were delivered by companies without the depth to run them, a restriction the government moved to ease for major projects after the outage. Contractors rotate off when a project closes. No durable engineering team owns any of it across its life.

There is a second limit that outages make invisible, because it is continuous. The National Information Society Agency’s annual survey of the digital divide has consistently found older adults, low-income households and people with disabilities operating well below the general population’s level of digital capability. When a service is available online it tends, in practice, to become available primarily online — the queue moves to the terminal, the staff move to the phones, and the person who cannot use either is left with a legally guaranteed offline option that is slower every year. Korea’s rapid ageing makes that population larger, not smaller.

Neither limit is an argument against what was built. They are arguments about what the completeness of the build implies. The more thoroughly a state consolidates its interactions with citizens into one rail, the more a single component’s failure becomes the failure of the state’s front counter — and redundancy for an event that has not happened yet is always the first line cut from a tender.