Twenty-Two Years, Seventeen Data Points, and a File That Speaks With One Voice

Two in the morning in Hamburg. I open a file labelled football. Inside there is no match. No club, no player, no expected-goals figure, no pass. There is a woman, a man, a video that has existed for twenty-two years, and a filing waiting before a judge in Mexico City. I count again. Seventeen information points in the record. Fourteen cite no source at all. What remains, no more than four points, all flows through a single door: counsel for the complainant. There are numbers that only tell the truth at midnight. Fourteen out of seventeen is one of them. It says nothing about football. It says something about how a case file is assembled, and how readily we mistake a brief for a record of fact. The complainant is Michelle Vieth, an actress. The respondent is Héctor Soberón, an actor. The filing was lodged before a Control Judge — Juez de Control — of the Superior Court of Justice of Mexico City. The content: allegations of gender violence, digital violence and media violence. The specific request: protection measures. The platforms named in the statement: Instagram, Facebook, YouTube. Counsel for the complainant: Jauregui Abogados. One thing must be said at once and said plainly: everything I have is allegation, not ruling. The record contains no response of any kind from Héctor Soberón or his counsel. No statement, no denial, no argument. That means anything I write below about the respondent is provisional and structurally incomplete. Time is the axis of the whole story. The video dates back twenty-two years. But the filing does not sue the dissemination of twenty-two years ago. It sues recent public statements, and argues that those statements restarted the circulation of the material, producing a harm that is ongoing rather than closed. Counsel for the complainant calls this revictimisation. They have also called on media and citizens to stop searching for and disseminating the material. The complainant has chosen public silence while proceedings run. The judge has not ruled. And alongside the allegations there is a threshold fact in factual dispute; the record does not establish it, so I will not speculate about it. In almost every legal system, conduct that occurred twenty-two years ago runs into a limitation barrier. That is why the architecture of this filing deserves closer reading than the content of the allegations. It spends no energy proving the old conduct. It spends its energy proving that there is new conduct, happening now, and that the new conduct is the recent statements. This is a coherent strategic choice. It is also the most fragile point. If the court accepts the framing, the file opens a precedent worth far more than the two people involved: harm could be re-established every time the old material is mentioned. If the court rejects it, the core of the structure collapses, because the underlying conduct is time-barred. But there is a gap in the chain of reasoning that the record cannot fill. The argument holds that recent statements revived searching and circulation. That is a plausible hypothesis. It is not a demonstrated causal chain. The record contains no search-traffic data, no before-and-after comparison, no cross-referenced timeline. A correlation presented as a cause — and this is the oldest trap in my trade. Correlation is not causation. It is only an invitation to go looking for causation. Then comes the sourcing audit, the part that kept me at the desk longest. Seventeen information points. Fourteen record their source as none. The remainder cite sources — and every source cited, without a single exception, is a statement by the complainant herself or by counsel for the complainant. Not one court document. Not one response from the other side. Not one independent source. I do not call that a lie. I call it a brief. The distinction is large. A brief can be entirely honest to the person who wrote it and still not be a record of fact. It is designed to persuade, not to balance. And when a file has only one door, there is no self-correction mechanism inside it — no second voice to question, to rebut, or simply to blur a detail that fits too neatly. Data is a temple, and I am only the one sweeping the leaves. But the one sweeping the leaves has to know which door he is sweeping at. Alongside that sits an asymmetry of public standing. The complainant is in what I call a high-presence, low-agency position. Her name is repeated widely, but her capacity for public action is locked by the logic of the case itself: on the theory her own side advances, every time she speaks the circulation restarts. Silence, then, is not weakness. It is a strategy consistent with the argument. On the respondent's side I cannot assess anything. No statement, no denial, no argument appears in the record. That absence is not evidence of guilt. It is a gap in the data, and a gap must be recorded as a gap. One further detail is worth noting: the judicial authority itself has been placed under public expectation. The record quotes a call for the court to rule with a gender perspective and in adherence to law. That phrasing cuts both ways: it raises the transparency bar, and it pre-commits one side to reading an unfavourable ruling as a procedural question rather than a substantive one. There is a technical detail worth flagging about how the file reached the outside world. All information came from a single law firm through a single communication channel. That means high message discipline and no internal contradictions among the information points — but it also means no second voice on the same side for cross-checking. A file with no self-correction mechanism on either side is a file where the reader has to bring their own ruler. And there is a purely data-level problem that twenty-two years creates: evidence decay. The original record of a video from two decades ago is rarely intact in format, metadata, or first point of dissemination. What survives is usually a copy of a copy. In my trade we call that a broken chain of provenance. When provenance breaks, everything downstream becomes inference. And finally, the thing that had me opening this file at two in the morning: the domain label. The record is labelled football. Not one club, not one player, not one metric across all seventeen information points. The wrong label is not a small error. It is evidence that the ingestion layer does not verify content against the label assigned to it. If that fault is systemic, other records in the same batch may carry the same defect — and a system quietly mixing legal data into a football entity graph will not raise an error. It simply erodes the reliability of the whole dataset, one record at a time. Stand far enough back and every heatmap becomes a painting. Stand close enough and a seventeen-point record becomes one too: the left half bright because someone aimed a light at it, the right half dark because nobody switched a light on. The biggest risk in this file is not legal. It is informational. A live, unresolved proceeding. A single organised source with high message discipline. A completely wrong domain label. Together those four produce a high risk rating — but a risk rating about data quality, not a comment on who is right. I have to be explicit here, because this is where my profession is most often misread: assessing the reliability of a record is not assessing its content. My model collapsed. I did not. I was still carrying an old assumption — that a legal filing implies a documentary record alongside it. Tonight that assumption broke. There is a filing, and there is a summary written by one side. What would change my assessment? Not a stronger statement from the complainant's side. Three very specific things: a copy of a court document with a case number; a formal response from the respondent; and a traffic series showing the search peak arriving after the statements, not before. Without those three, any conclusion is just a hypothesis presented neatly. And there is a paradox I owe the reader before I close: writing about this case is itself an act that raises the chance others find it. Every article about a video is one more small indexing of that video. The complainant's side understands this well enough to have built it into their own argument. Which means that I, sitting here writing, am inside the variable they are trying to control. The only decent way I can behave is to describe no content, publish no identifying detail, and refuse to turn a harm into a story to read at two in the morning. The signals for the next cycle are clear and trackable. First, the scope of the protection measures when the judge rules. A narrow order — barring repetition of a few specific statements — is a categorically different outcome from a broad order barring all commentary on the case. Second, whether the ruling confronts the temporal question directly or sidesteps it. Third, whether the respondent files a response; the day he speaks is the day this record stops being one-sided. What I take from tonight is not a conclusion about who is right. I have no data for that conclusion, and a person without data has to keep quiet in the right places. What I take is a question I will ask of every record from now on: does its label match its content? Two in the morning. I close the file and write one line in my notebook: not football.