Falsification Register

By 31 December 2028, on the Ministry of Justice's published family court statistics, the proportion of private law children cases involving alleged domestic abuse in which unsupervised contact is nonetheless ordered will not have fallen by more than a fifth from its 2025 level, unless the costs regime has changed so that a contested hearing costs the party who contests it more than a settled one.

The proportion falling further than a fifth while the incentives stand as they stand now.

The conventional account says an institution is fixed by replacing the people at the top. This shows in two arenas and both must hold.

In institutions. An institution that answers a scandal by changing its people but not its structure is recaptured. Change the structure and it holds, even if the new people are no better. The Post Office, emerging from the Horizon inquiry into the wrongful prosecution of hundreds of subpostmasters on faulty accounting software, is the live test. The inquiry recommended a standing independent body to deliver redress. The government has not built it. The framework predicts the adversarial conduct towards postmasters persists while the architecture is unchanged.

In the machines. Models trained against an explicit constitution will refuse harmful instrumental actions, deception, sabotage and the targeting of the vulnerable, at higher rates than equally capable models without one, and the gap will widen as capability grows rather than close.

In institutions: the adversarial conduct falling away while the architecture stands as it stands now. In the machines: a more capable model built without that architecture proving just as safe.

The conventional account says the problem is bad lawyers and under-resourced courts. Where a settlement is reached after months of billable negotiation and then collapses on the day, the manoeuvre described in Chapter 17 is structural, not incidental. Reforms that add pattern-evidence and penalise strategic delay will reduce it. Courts that leave the incentives untouched will not improve.

The manoeuvre falling away where the incentives are unchanged, or persisting where they have been changed. Changed means a named reform, published, with a date on it.

The conventional account says the platforms will clean up reputation in the end, because a market in trust is worth more to them than a market in lies. As synthetic media scales, manufactured reputation increasingly drowns authentic testimony, so the next victim arrives to find the predator's curated record already in place.

Reputation recovering as a trustworthy signal through the platforms' own incentives, with no containment architecture for provenance built.

Fix the incentives in a system and most of the harm goes away, because most of the people were responding to the incentives. That much is undisputed. This prediction is about what is left. Architecture contains the behaviour of the many. It does not change the wiring, and the wiring keeps turning up.

The worst of it will stop falling and sit there. When you look at who is holding it up, it will be a small number of the same people, turning up somewhere else, doing the same thing in a different building under different rules. The pattern follows the person, not the room.

Reform a system through consequences and measure it honestly. If the worst of the harm falls as far as the ordinary harm falls, the wiring described in the book is doing no work and the central claim is wrong.

Staked in Chapter 25. A laboratory built as the foundation laboratories are built, the chief executive chairing the board that oversees the chief executive, the safety team reporting up the chain it exists to check, the outside scrutiny funded and helped into design by the industry it scrutinises, will have its safety function hollowed, overridden or quietly rerouted within the decade, whoever runs it and however good their intentions.

A laboratory so built keeping its safety function intact, independent and effective through the decade. Losing it means the framework is wrong on this point.

If the argument is right, what is left over when you have fixed the rules is people, and people move. Every public inquiry in this country ends by naming the senior people who let it happen. Grenfell named them. The Post Office named them. Take that list. Then build a second list next to it: people at the same level, in the same industry, for the same number of years, in organisations that were looked at and found to be sound. Now go back through the public record and count, on both lists, how many of those people had already been found at fault somewhere else: a different employer, a different sector, a different decade, nothing to do with the case in hand.

The prediction, fixed now so that the line cannot be moved afterwards: the first list has at least twice as many.

Being caught once makes you easier to catch twice, so whoever runs this counts only the cases that surfaced on their own, and the person doing the counting does not know which list they are working from. If that cannot be done, the test is no good, and this register will say so.

The two lists coming out the same. Then the people were interchangeable, the building did it and not the person standing in it, and the argument of the book is wrong. Not partly wrong. Wrong.