The Post Office treated Horizon's data as always accurate

A shortfall on a branch account is a number a person has to explain. If the institution has already decided the system that produced the number cannot be wrong, the explanation left to the person in the branch is a confession. Sir Wyn Williams puts that stance, and what followed from it, at the front of his report.

Volume 1 of the Post Office Horizon IT Inquiry was presented to Parliament on 8 July 2025. It is HC 1119. He considered evidence and information available to him up to and including 6 June 2025. He published this volume when it was finished, rather than waiting for the rest of the report, because he wanted the human impact in front of the Minister first. Later volumes are where he says he will deal more fully with the behaviour of people at the Post Office and at Fujitsu. This volume already states what he is satisfied of, and it is enough for a people leader to use.

Post Office Counters Limited began rolling Horizon out to branches and Crown Offices in about September 1999. By about the end of 2001 or early 2002 the rollout was complete, and from then on every branch and Crown Office has used a version of Horizon. The first version, Legacy Horizon, was developed for and supplied by ICL Pathway Limited, and it stayed in use until 2010. Before rollout, some Fujitsu employees had found that Legacy Horizon could produce data which was false. From time to time it was hit by what he calls bugs, errors and defects, and those could show losses or gains in a branch account that were illusory rather than real.

He is satisfied that a number of senior employees of the Post Office, and employees who were not senior, knew or at the very least should have known that Legacy Horizon could do this. Many of the people who gave evidence were very reluctant to accept it. For all practical purposes, throughout the life of Legacy Horizon, the Post Office maintained what he calls the fiction that its data was always accurate.

In 2010 Legacy Horizon was replaced by HNG-X, also called Horizon Online, developed for and supplied by Fujitsu. It too was afflicted, from time to time, by bugs, errors and defects that showed illusory gains and losses. He is satisfied that a number of employees of Fujitsu and of the Post Office knew that. In 2017 Horizon Online was replaced by HNG-A, again from Fujitsu, and that version, with some later modifications, was in use when he began hearing evidence. He had been working on the assumption that this version was far stronger than earlier Horizon, or at least different and stronger. Evidence from Fujitsu, and from postmasters who use it, means that assumption may no longer be wholly justified. He leaves the explanation to a later volume.

Between 2000 and the autumn of 2013 the Post Office prosecuted postmasters and others who worked in branches and Crown Offices in England and Wales, relying on Horizon data to prove that real losses had occurred and that the explanation was theft, false accounting or fraud. In Northern Ireland and Scotland the prosecuting authorities brought dishonesty cases over the same period and, in his account, for some years beyond 2013. In each of those cases, or at least in most of them, Horizon data was used to prove the losses were real, and the Post Office, the prosecuting authorities, or both asserted that the data was wholly reliable. The same assertion supported holding postmasters to their contracts for apparent losses.

The consequence, as he states it, is that many hundreds of people were wrongly convicted of criminal offences, and many thousands were wrongly held responsible for losses that were illusory. He says all of them are properly regarded as victims of wholly unacceptable behaviour by a number of individuals employed by or associated with the Post Office and Fujitsu, and by those institutions. The impact on people wrongly accused and prosecuted has been disastrous. The impact on people held liable for losses that had not occurred has also been disastrous in many instances. Close family members suffered too.

Complaints started almost as soon as Legacy Horizon was rolled out. Postmasters said Horizon was, on occasions, producing false data that affected branch accounts. The complaints continued through Legacy Horizon and Horizon Online, and they grew louder. Members of Parliament supported the postmasters. The Post Office, in his words, trenchantly resisted the contention that Horizon sometimes produced false data. In 2009 the Justice for Subpostmasters Alliance was formed. By 2016 it was ready for a court fight. About 555 claimants were enough for a Group Litigation Order. The case was hard fought. It ended after Mr Justice Fraser produced two substantive judgments in which the claimants' main contentions were accepted without reservation.

He will not put a precise count on everyone who suffered from the misplaced reliance on Horizon data. He does say, with a degree of confidence, that there are currently about 10,000 eligible claimants in the redress schemes, and that the number is likely to rise by at least hundreds. The range runs from people held liable for small sums, perhaps tens or hundreds of pounds, to people wrongly convicted and imprisoned, people who became seriously ill, and people who were made bankrupt. Some claimants represent people who have died: some of natural causes before a conviction was quashed, some more recently before a claim was fully determined, and some who are said to have been driven to despair and suicide. In the case illustrations he is careful. They are drawn mainly from Phase 1 evidence, and he does not, at this stage, make factual findings about precisely what occurred in each individual case. The focus of that section is the human impact said to have followed the discovery of a Horizon shortfall.

The redress chapters are the part a sponsor can misread as administration. Postmasters have described delay, a lack of information, and settling for less than they believe is full and fair. Some older postmasters fear they will have little time to benefit from whatever arrives. For some, the application itself compounds the harm. He records an Advisory Board view that the schemes needed a culture change, away from a strictly legalistic assessment and toward what the Board called a more flexible and compassionate approach, in keeping with ministers' stated aim that redress be full and fair. His own recommendations, nineteen of them, ask for written responses by 10 October 2025 from government and, where appropriate, the Post Office and Fujitsu. They ask the Minister and the Department, with the Post Office, to say in public what full and fair financial redress means: sums equivalent to the top of the range a judge in England and Wales would award in civil litigation, with room to depart from ordinary principles if fairness in a particular case demands it. They ask for a standing public body that can devise and deliver redress when a public body has wronged people. They ask the Department, Fujitsu and the Post Office to publish, by 31 October 2025, any agreed programme of restorative justice, or the actions taken toward one. Fujitsu, in that recommendation, means both Fujitsu Services Limited and Fujitsu Limited. A scheme that exists is not, on this volume, the end of the duty.

The question I would want on the table, before anyone sanctions the person a system has named, is the one this volume answers late. Who knew the number could be false, and what did the organisation do with that knowledge? On his findings, people inside the Post Office knew or should have known, Fujitsu employees had seen false data before rollout, and the institution still maintained that the data was always accurate. Prosecutions and contract demands proceeded on that assertion. The person closest to the branch was the one left to answer for an illusory loss.

A model, a dashboard or a case system that produces a figure a person must defend sits in the same structural spot, even when the stakes are a warning, a performance mark or a payment stopped, rather than a prosecution. The vendor's assurance that the system is reliable is not a substitute for the question he asked of Horizon. The question is what the organisation does in the hour after someone who uses it says the figure is false.

1. Separate the claim that the system is reliable from the person the system accuses. The owner of the reliability claim has to be able to produce the known defects. In this report, the defects had names. Bugs, errors and defects could show a loss that was not real, and employees knew or should have known. A pack that still says the output is always accurate has already chosen a side.

2. Take the first complaint as a test of the system. Complaints began almost as soon as Legacy Horizon was in branches, and they continued for years. The response he records is resistance, including while members of Parliament were involved. A queue of tickets that never reopens the reliability claim is that response in a calmer office.

3. Do not close the matter because a scheme, a policy or a vendor note exists. About 10,000 people are in the redress schemes. He still finds delay, missing information, and offers people do not regard as full and fair, and he asks for a public definition of that phrase, a standing body, and a restorative programme. A page that says the issue is being handled is not the same as a decision someone can inspect.

4. Write down the next hour. Who arrives when a user says the output is false. What they are allowed to stop. What they are not allowed to do to the person who raised it. Thanking that person, and holding the sanction, is the minimum this history makes obvious. The prosecutions and the contract demands are what that hour looked like when it was missing.

A psychological safety session can be where an intact team takes one live output, a score, a shortfall or a recommendation, and agrees that hour in writing before the next time the system names a person.