It had a real promise of operation benefit when first rolled out throughout the Post Office's network of about 18,000 branches in 1999. The principal institutional gain was the Post Office itself: a uniform, computerised accounting system that superseded a variety of ad hoc-ledger systems, provided a uniform system of record keeping and, for head office, a single view of financial performance across the whole branch network. This degree of visibility allowed for identifying genuine fraud more easily, as well as identifying underperforming branches and managing cash reconciliation at scale.

As long-term developer and maintenance contractor, Fujitsu reaped the profits from a government contract for decades – despite all of the evidence that the system failed them. This financial relationship itself had ethical implications: the contractor was on a commercial stake in keeping the impression that Horizon was working properly.

The invisible advantage for the customers of the Post Office was the speed and uniformity of the transactions carried out, and for the general public was the modernisation of a national institution that was starting to lose a foothold with competition from high street banks and online services. As far as the technology was concerned, in a narrow operational sense, it brought some of what it promised.

What matters is that there was not equality in the benefits and that there was an underlying unreliability that was kept hidden. The operational benefits of the system were claimed by the institution and the contractor and the expenses of its failure were shifted, unnoticed and unarticulated, to the least able to carry them.

The broader public

• • Local services were lost from communities due to removal of branch operators or branch closure as a result of prosecutions.

  • The public's confidence in digital systems was significantly damaged; and in government institutions there was a harm to public confidence, which is diffuse but material and long-term.

  • Anyone subjected to automated decision-making in other settings and future claimants were left with a legal landscape where Computer Evidence was admitted against them with limited judicial review and the Horizon precedent had set a precedent for such practice.

Ethical Consequences

The Horizon scandal is a multi-level ethical debacle. The Post Office was wrong in failing to exercise good faith towards postmasters. Prosecutors failed to make disclosures. The Fujitsu engineers who knew of the known bugs failed to get the information to those whose lives depended on it. Digital evidence was not adequately investigated by courts. One injustice added to another, making injustice a structural possibility rather than an inevitability.

The case also revealed the ethical shortcomings of using a legal presumption (computer records are accurate unless challenged) in lieu of true scrutiny. The ease it was for both courts and prosecuting organisations to presume the accuracy of the computer was a benefit, because it effectively shifted the burden of proof to the people lacking the technical expertise or financial resources to test it.

The biggest long-term legal impact is that of the reform of the legal presumption of computer reliability in English law. The Law Commission has suggested that the presumption should be reversed to make it more difficult for parties wanting to use computer generated evidence in criminal cases. The implications of this reform, which was directly sparked by the scandal of Horizon, will be long lasting in the ways in which digital evidence is viewed in court.

It has hastened UK public policy evolution of algorithmic accountability frameworks. It has helped to shape the conversation on the use of automated systems in benefits assessment (Universal Credit), immigration enforcement, and social care, all areas where similar power dynamics are found between the state and people who are subject to automated decisions.

Scandal has changed a generation of British citizens' perceptions of institutional control and digital processes at a cultural level, in a way that will be lasting. The idea that big organisations and the technology they use are to be trusted without question has been undermined - an awkward corrective in a society that is increasingly run by algorithmic processes.

The case has become a benchmark globally for similar scandals involving the use of automated algorithms in the collection of debt and the identification of fraud, such as the Australian government's Rob debt disaster, where automated debt recovery notices were sent to welfare recipients based on algorithms that were overly susceptible to fraud; the Dutch tax authority's algorithmic fraud detection system, which failed to accurately identify fraud and targeted families of ethnic minority backgrounds; and a range of different jurisdictions in the United States that have implemented risk assessment tools in criminal sentencing. They share a feature in common with Horizon: they are both structurally perverse, involving automatic authority, hidden evidence of wrong, and focused harm on a symptomatically weak population.