Society & PolicySubscribers only yesterday8Add to bookmarks

The Verge publishes an immersion into a convention "the future of policing in the digital age" in Fort Worth. Verdict: AI is no longer entering policing through the pilot project door - it's entering through the vendor catalog door.
We no longer ask if AI is used by the police - we ask who they buy it from. A trade show in Fort Worth exposes the mechanics: vendors, demos, contracts, everything is in place, without a single audit standard having been adopted.
On July 16, 2026, The Verge publishes a lengthy investigation, "COMPUTER COPS: Inside the big business of selling AI to the police," signed from Fort Worth, Texas. The press was excluded - the author reconstructs the day from accounts of participants gathered outside the site. It's already a signal: this market prefers to sell itself to its institutional buyers without external witnesses.
What The Verge reports is the shift from a phase of "isolated tools" (facial recognition, ANPR, flow analysis) to a platform phase: integrated suites that, according to the article, "AI is threatening to seize the very heart of policing." The buyers are police chiefs and deputies; the sellers, a constellation where the entry of smaller suppliers creates a commodity effect - police AI becomes a catalog product, not a project.
Three plumbing components make up the typical offering observed at this type of trade show: a municipal data lake (cameras, police calls, files); a model layer that produces alerts and incident summaries; a dashboard turned towards the officer. The weak point of governance is the transition from the second to the third: the way an "alert" produced by a model becomes a field operational order is almost never traced in contracts, and rarely in internal procedures.
This is not a technical question, it's a question of sequence. Each contract signed today engraves a supplier, a data pipeline, and a model of responsibility into the daily operation of a local police force - long before the audit frameworks that should govern it are adopted. The reader who decides to open this market - elected official, prosecutor, CIO of the agglomeration - inherits an integration that is difficult to undo. To follow: which states will adopt a targeted moratorium, as New York did for data centers, but on predictive policing contracts.
Create a free account to access all our content and the weekly review.
Article produced by artificial intelligence, reviewed under human editorial control.
Sign in to join the discussion.
L'IA dans la police, ça peut être utile, mais comment être sûr qu'elle ne servira pas à opprimer ?
L'IA dans la police, pourquoi pas, mais à condition que ce soit transparent et contrôlé.
L'absence de contrôle sur l'IA policière m'inquiète. Qui vérifie que ces outils sont utilisés de manière éthique ?
L'IA saura-t-elle vraiment saisir le contexte et les intentions humaines dans la police ?
L'IA va-t-elle amplifier les biais des forces de l'ordre ou les réduire ?
Comment garantir que l'IA ne porte pas atteinte à nos libertés ?
L'IA pourra-t-elle vraiment saisir les subtilités du comportement humain ?
Et l'éthique dans tout ça ? L'IA peut-elle vraiment remplacer le jugement humain des policiers ?