Fujitsu has paid nothing
And other things the procurement paperwork confessed in November.
In May 2009, Computer Weekly published a feature about seven sub-postmasters who ran branches of Britain’s Post Office and had been bankrupted, prosecuted or driven to despair by an accounting system that did not work. The piece was written by Rebecca Thomson after a year of investigation. It told the stories of Lee Castleton, Jo Hamilton and five others. It went, as Thomson later put it to The Times, out to ‘a clanging silence’.
A year later Thomson left the magazine and a colleague named Karl Flinders took the file. He has stayed on it for fifteen years. By his own count, given to Press Gazette in January 2024, Computer Weekly published roughly 350 stories on the Horizon scandal between Thomson’s first piece and the broadcast of the ITV drama Mr Bates vs The Post Office. Around seventy of those landed before the end of 2018; the rest came in the five years that followed as the group litigation, the Court of Appeal judgments, the public inquiry and the drama produced successive waves of news. The reporting drew in colleagues, but the through-line is Flinders. He was working with the Justice for Sub-postmasters Alliance, which Alan Bates had founded in direct response to Thomson’s 2009 piece. The two were necessary to each other.
Around them, Private Eye carried it from 2011 onwards, with Richard Brooks doing the work, joined eventually by the freelance Nick Wallis. Wallis was the human bridge between the trade press, the BBC and the eventual ITV drama. He produced three Panoramas, made the Radio 4 series The Great Post Office Trial, ran his own crowdfunded site through the Bates v Post Office trial, wrote the book and was series consultant on the drama. He was never on staff anywhere. His bills were paid, in part, by Private Eye and by his readers.
The BBC’s place in this picture matters and is easy to misread. Three Panoramas and a Radio 4 series sound, on the face of it, like institutional commitment. The reality is closer to hospitality. Wallis was the freelance who brought the story; the BBC provided the venues. The corporation’s news operation did not staff Horizon as a running political scandal until the ITV drama forced it to and there is on the public record evidence that the Post Office’s legal pressure on the corporation slowed the pre-2024 work in ways the published Panoramas do not show. The BBC was a venue. It was not a champion. The distinction matters because the absence of an institutional champion is what made the fifteen-year defensive period possible.
This is a kind of journalism. It is not the kind that runs as a front page in a national newspaper.
The print nationals were not silent on Horizon between 2009 and 2024. The Times, the Telegraph, the Mail, the Guardian and the Financial Times all ran pieces, sometimes substantial, around court dates, parliamentary debates and Computer Weekly disclosures. Neil Tweedie wrote a long Daily Mail feature in the early 2010s. The Guardian carried the 2018 and 2019 group litigation. What they did not do was staff the story. There was no Horizon correspondent at any print national. Coverage was reactive, court-date driven and Computer-Weekly-derived. As Flinders told Press Gazette: ‘The mainstream media was a bit slow to it because it was a bit risky and they probably thought it was a bit boring. Computers and the Post Office.’
The clearest test is the judgment of 16 December 2019. Mr Justice Fraser’s 313-page ruling in Bates v Post Office found Horizon unreliable and described the Post Office’s defence as ‘the 21st century equivalent of maintaining that the Earth is flat’. This was, on any honest news judgment, a front page. It did not run as a front page in any national paper. The Court of Appeal followed in April 2021, quashing 39 convictions and calling the prosecutions ‘an affront to justice’. Three further years passed before Parliament legislated mass exoneration. What forced the legislation was not the Court of Appeal. It was a four-part teatime drama on ITV that ran on New Year’s Day 2024, watched by 7.3 million people on linear broadcast. Within a week, every print national had a Horizon correspondent. The story they had been writing about for fifteen years became a daily political scandal in seven days.
That gap - between a senior judge using the language of flat-earthism in a public judgment in 2019 and Parliament acting in 2024 only after the story had been dramatised - is the most damning thing the British print press has done with a Court of Appeal judgment in the last decade. It is not that the nationals failed to cover Horizon. It is that they failed to make it the running political story that the documents had made it from December 2019 onwards.
Then November 2025.
Karl Flinders, still at Computer Weekly, reported that the Post Office had quietly extended Fujitsu’s Horizon contract by another year, without a competitive tender. The procurement notice, published by Post Office Limited and available on the public commercial register, set the value at £49.2 million inclusive of VAT. PublicTechnology picked the story up a week later. The notice itself contained, in language unusually plain for Whitehall, an institutional confession.
‘The existing Horizon system,’ it said, ‘is a highly complex legacy platform, written in outdated versions of software languages and incorporates five “systems” in one: financial services; banking; government services; mails; and retail. Horizon is an aging platform and has an inflexible monolithic architecture that makes technology change difficult. It would not be possible to simply hand this system over to another supplier.’
The state that prosecuted around nine hundred of its own sub-postmasters on evidence produced by Horizon had now formally conceded, in a public commercial document, that it built something it could not unbuild. The system stays. The vendor stays. The previous bridging extension, signed in December 2024, was eventually valued at £75 million. November’s was £49 million. The cumulative additional spend on the two extensions has passed £124 million. Fujitsu has paid nothing into the £2 billion redress regime running alongside.
A government spokesperson explained the decision in language that read as architectural confession at one further remove. ‘The fact they still use the Horizon system indicates past under-investment, which can’t be rectified overnight.’ That is not a justification. That is an admission.
A serious search of the public web for substantive coverage of the November 2025 procurement notice in the Times, the Telegraph, the Guardian, the Financial Times, the Mail or the BBC returns no direct hits. The story remained, between November 2025 and the end of April 2026, in the trade and specialist press: Computer Weekly, PublicTechnology, The Register. The £49.2 million figure, the ‘would not be possible’ phrasing, the political resonance with Sir David Davis’s parliamentary calls to use new procurement powers to debar suppliers with terrible track records - none of this surfaced in a national piece in the four months that followed.
Four months later, on 13 March 2026, the Business and Trade Committee published a report that included Liam Byrne’s line that ‘Fujitsu has yet to contribute a penny to the nearly £2 billion redress bill, even as it continues to benefit from public contracts.’ Byrne went on Today that morning. The Press Association ran the wire. ITV News carried it. The nationals picked it up. The procurement notice that was the underlying fact of Byrne’s complaint - the contract that had just been quietly extended for £49 million - was not reattached to the story even on the same news cycle. The select committee shouted; the wire ran; the procurement document stayed in the trade press where it had been filed.
The familiar reading of these cases is that automation went wrong. Three states, three automated systems, three failures.
The reading does not survive contact with the documents.
In the Netherlands between roughly 2005 and 2019, the Dutch tax authority’s childcare benefits division accused tens of thousands of families of fraud they had not committed. The flagging algorithm scored claims partly on the basis of nationality. The recovery regime, upheld by the Council of State for over a decade, demanded full repayment for any administrative shortfall. Wages were garnished. Assets were seized. The pattern story was broken in 2018 by Pieter Klein at RTL Nieuws and Jan Kleinnijenhuis at Trouw, a national daily that followed the story where its larger competitors would not. De Volkskrant‘s editor-in-chief later acknowledged on Dutch public television that his paper had received a tip about the affair as early as 2017 and had not picked it up. The Rutte government fell on 15 January 2021. No civil servant or minister has been criminally prosecuted. State immunity holds.
In Australia between 2016 and 2019, the welfare agency raised around 470,000 debts against benefit recipients on the basis of an income-averaging formula that internal legal advice had warned, two years before launch, did not accord with legislation. The advice was suppressed. Christopher Knaus broke the story in Guardian Australia on 23 December 2016, three days before Christmas, off a whistleblower tip. Luke Henriques-Gomes joined as welfare and inequality reporter in 2018 and carried it sustainedly through the federal court challenge, the unlawfulness finding, the abandonment of the scheme, the Royal Commission and the report. They did the work. Rick Morton, then at News Corp’s The Australian, has since said in print that he ‘mucked up’ early Robodebt stories and steered clear, partly because of an editorial culture in which ‘welfare crackdown’ stories were favoured. Two suicides are named in the Royal Commission record. On 11 March this year, almost ten years after the scheme launched, Australia’s National Anti-Corruption Commission published findings of serious corrupt conduct against two career officials. The four ministers responsible, including a former prime minister, were cleared.
The cases share an architecture. In each, the institution presumed the system’s output was correct and the citizen’s denial suspect. The presumption was made operationally effective by limiting the citizen’s access to the underlying data. Sub-postmasters could not interrogate Horizon. Dutch parents could not see the risk indicators that had flagged them. Australian recipients could not see the tax-office data the debts were calculated against. The architecture made the presumption not a question of attitude but a question of access. You cannot defend yourself against evidence you are not allowed to see.
They share a consequence. The citizens who were wronged went to prison, lost their children, lost their homes and in documented cases lost their lives. Fujitsu has paid nothing and continues to win billions in further public contracts. The Dutch tax authority was fined €6.4 million; the fine was paid out of one part of the Dutch government and into another. The Australian career officials found by the corruption commission to have engaged in serious corrupt conduct had already left the public service and could not be sanctioned. Citizens pay in years. Institutions pay in headlines.
They share a pattern of reporting. In each case, sustained attention was paid for years by specialist or single-desk journalists working largely without the support of their wider national press. Computer Weekly carried Horizon for fifteen years. Klein and Kleinnijenhuis carried the toeslagenaffaire while De Volkskrant sat on the same tip. Knaus and Henriques-Gomes carried Robodebt at Guardian Australia while News Corp titles ran the government’s framing. In each case, the political response only arrived when an external trigger forced it: an ITV drama, a parliamentary inquiry, a change of government.
This is not a complaint about journalism. It is a structural observation about it.
The print national news desk is wired for political prominence, dramatic narrative and source novelty. These are the incentives the form was designed to reward. They produce excellent journalism in their right register. Andrew Norfolk’s work on Rotherham at The Times, Carole Cadwalladr on Cambridge Analytica at the Observer, Amelia Gentleman on Windrush at the Guardian - all sustained, all consequential, all done by single staff investigators at print nationals over months and years. Norfolk’s work is the strongest example. He carried Rotherham child sexual exploitation from 2011 to 2014, with editors who backed him, deep pockets to defend the High Court injunction the Council attempted in 2012 and a political and policy response that arrived within his publishing window. The print nationals can do this work. They have done it.
The combination Norfolk had - a committed staff investigator, willing senior editors, the resources to defend a legal challenge and a high-profile antagonist in the abusers themselves - is rarer than the cases collected here suggest it should be. Horizon never had any of it. There was no dramatic event the picture desk could photograph, no named villain, no quotable judge until 2019, just trade-press documentation accumulating in silence. The story did not fit the form.
The trade and specialist press is wired for institutional complexity. Its readership is the institution being reported on, which means the form rewards forensic, technical, procedurally accurate, sustained attention. Computer Weekly’s readers are people who buy and run computer systems. Inside Housing’s readers are people who manage social housing. Their reporters do not need to find a quotable judge or a named whistleblower or a high-profile antagonist for a story to land. They can publish on the seventeenth iteration of a contract dispute and their readers will read it.
The gap between the two is where citizens fall. When an institutional failure is procedurally complex, low-profile and slow-moving - when the antagonist is the Post Office rather than Facebook - the trade press attends to it from early disclosure and the print nationals do not staff it until something dramatic happens. The defensive period of the institution and the defensive period of the press run in parallel. They are the same period.
The procurement notice published in November is the architecture of this gap, written down. It records, in the institution’s own commercial paperwork, that the British state cannot extricate itself from the vendor whose product produced the most widespread miscarriage of justice in British history. Karl Flinders read the document. PublicTechnology read it. The print nationals did not read it and four months later, when a parliamentary committee shouted about the same underlying fact, the document was still not re-read.
The pattern is not that states sometimes get automation wrong. The pattern is that states defend automated wrong for as long as they are allowed to and that the political response arrives only when the press allows it to. Some kinds of press will allow it sooner than others. Knowing the difference is itself a kind of literacy. It is what this Substack is trying to teach.
When you next see a story called a scandal that has just broken, ask how long the trade press has been on it.
True Regard is the practice of reading the architecture across the cases. The book that lays it out, The Extraction Pattern, publishes in June. The Substack runs alongside.