OPINION: Anne-Sofie Engelschiøn
Two stories, two countries, same failure. Most people have never heard of either, and that alone should tell us something about how quietly these things happen.
In the UK, between 1999 and 2015, the Post Office prosecuted hundreds of ordinary sub postmasters for theft and fraud. The evidence came from Horizon, an accounting system built by Fujitsu that had bugs nobody wanted to admit to. Over 900 people were wrongfully convicted. People lost their homes, their marriages, their freedom. Some took their own lives. The Post Office had the unusual power to investigate and prosecute its own cases, and for two decades it used that power to protect the system’s reputation instead of questioning the system itself.
In Norway, NAV, the welfare agency, misread EEA rules for years and told people they had to stay physically inside the country to keep benefits they were legally entitled to while traveling. More than 80 people were wrongly convicted, some imprisoned, and over 7,500 were affected in some way. The Blindsonen report that followed found something almost more unsettling than malice: NAV, the police, the prosecutors, and the courts had all simply trusted the same wrong interpretation, one after another, without a single one of them stopping to check it against the actual law.
What strikes me most, reading these two stories side by side, is not that mistakes happened. Mistakes always happen. What is missing in both cases is a checkpoint, a built in moment where someone with the authority to pause was required to ask whether the pattern in front of them meant the system itself might be wrong, rather than the person.
One of the problems was that they were treated as hundreds of isolated incidents, one at a time, for years. Nobody connected them. Nobody had the job of connecting them. In the Post Office’s case there was also a real incentive not to look too closely, because admitting Horizon was flawed meant admitting the company had ruined lives and would owe a reckoning it could not easily afford. In NAV’s case it was closer to institutional autopilot, a wrong answer that nobody downstream felt entitled to question because the system had already said it.
Any system that makes decisions about people, whether it is software, policy, or a set of internal rules, will eventually be wrong somewhere. The interesting part is whether the system has a place built into it where being wrong can surface early, before it becomes hundreds of ruined lives and a decade of appeals.
Years later, accountability in both cases is still unfinished. Fujitsu has yet to meaningfully contribute to compensation. Criminal investigations into the Post Office’s conduct are not expected to conclude before 2027 or 2028. In Norway, victims are still pursuing their case through the European Court of Human Rights. The people who built and ran these systems have mostly moved on. The people harmed by them are still waiting.
Is this simply what happens when humans build systems, because we are wired to trust structure over our own doubts and rarely stop to question something that looks official?
