Pauillac AI-CSAM Case Turns on Identifying Children, Not Counting Files

Searches of his home and studio produced what prosecutors described as CSAM on a massive scale—several thousand files—during only a partial examination of seized devices. The number is meaningful evidence of the alleged digital record, but it has no public denominator: it does not reveal the number of distinct children, the number of altered source photographs, or whether any individual local child appears in the material.

The accounts also leave the provenance of the alleged source images open. One report said the photographs were sometimes taken by the photographer himself. But another local account said the inquiry still had to determine whether the images and videos had been taken by the accused. That unresolved distinction is why a file count cannot be converted into a victim count.

Familiar access widened the possible scope

The photographer’s client work explains the scale of local anxiety, not the scope of confirmed harm. The same local account described work at schools, sports clubs, fairs, weddings, birthdays, baptisms, communions, galas and family, child and maternity sessions. His studio has since closed.

That record creates a broad pool of possible source images. It does not make everyone who used the studio a victim. In the later reporting, a parent group created to share information had more than 320 members; its membership measures concern, not exposure. Pauline and Karine, two mothers interviewed in the later investigation, said they did not know whether photographs of their children had been used. Both submitted information or photographs after the appeal.

The alleged technique is the key reason a routine photograph matters to the investigation. If an ordinary image was used as source material for an AI-generated sexual image, the underlying photograph would not itself have been sexual. But the public accounts do not disclose which files were altered, who appears in them, or whether any reported family’s photographs were involved.

An appeal for information is not a victim list

In December 2025, the Gironde gendarmerie asked people with useful information to come forward. It said that, under an investigating judge’s supervision, officers would directly contact the legal representatives of victims who were identified. The appeal also said an investigation was under way to identify victims.

The same account quoted the Gendarmerie as saying the alleged offending had affected many minors locally and had been stopped by the arrest. That is the force’s assessment, not a published identification list or a disclosed count. The public call is evidence that the identification work remained under way.

For families, submitting a photograph can help investigators test a possible link. It cannot by itself prove that the picture was altered or distributed. Nor does a lack of contact settle the opposite question while the matching work remains unfinished.

What evidence would resolve the central question

The next material evidence would be a judge-supervised match between seized files, any original photographs and identifiable children, followed by direct notification to their legal representatives. It would need to distinguish an image held on a device from an image that was AI-altered, identify the child depicted and establish the alleged origin and distribution of the material.

Until investigators publish such results or the court tests the evidence, the narrowest supported conclusion is also the most important: authorities allege an AI-enabled scheme and have reported several thousand files from a partial review, while the people potentially depicted have not been publicly counted or identified. The investigation’s outcome will turn on those matches, not on the file total alone.