Media & agencies
An hour of interview to pull three quotes, the same story rewritten for web, newsletter and vertical video, and a forty-year archive where nobody can find what ran in 2011. Inferana does that work on open models, inside the newsroom.
What each regulation in your sector requires, and which part of that the deployment solves.
Allows text and data mining unless the rightsholder has expressly reserved their rights.
Open models with published licence and model card. Your archive is never used to train anything.
Requires artificially generated or manipulated content to be labelled so the public can identify it.
An auditable record of every query: what was generated with AI, by whom and when.
Protects journalistic sources and the confidentiality of communications of those practising journalism.
Deployment inside the newsroom perimeter: notes from a conversation with a source never reach a third party.
Legal basis, minimisation and control over transfers in the processing of personal data.
No content logs by design: prompts and responses are never stored or used to train models.
The same ones you already use, specific to your sector.
Nobody has taken an edition off the calendar, but people have left the newsroom. What comes up in the morning meeting keeps repeating:
I record an hour of interview to get three quotes out of it.
NewsroomIt ran in 2011 and nobody remembers the headline it ran under.
Documentation & archiveI rewrite the same story four times: web, newsletter, vertical, social.
AudienceThe English edition ships a day late because we translate by hand.
International editionWe are half the size we were and we close the same pieces.
Deputy editorThe brief lands at seven and the proposal is presented on Monday.
Account directorNotes from a source ended up pasted into some chatbot.
Editorial directorWith clean audio the transcript is usable as it comes. With overlapping voices, room noise or a bad line it needs checking, especially names and figures. What you save either way is typing out the whole hour: you only check the timecode of the quotes that make the piece.
Anything that is text, or can be turned into text, including page-layout PDFs. Searches run on what a piece said, which is exactly where the CMS search gives up. The index sits in your own deployment and you take it with you if you leave.
You give it the style guide and a sample of published pieces and it works from that: use of acronyms, forms of address, headline length, how sources are attributed. What comes back is a draft to edit. The byline and the final read stay with whoever publishes.
With a deployment inside your own infrastructure the conversation never leaves the newsroom, and prompts are not stored. The editorial call on what material gets processed, and under which policy, stays with you, as with any other newsroom tool.
No. Not the archive, not the documents you index, not the prompts, not the responses. None of it enters any training run. What gets generated stays inside the deployment and you can export it whenever you want.
We do not give that guarantee and we do not offer indemnity against copyright claims, and be wary of anyone offering one without conditions. What you do get: open models with published licence and model card, so your legal counsel can assess each one, and editorial review that stays human.
Access control per user and per team, with a record of every query. One account's material is walled off from the rest, the same way an investigation in progress is walled off from the rest of the newsroom, and you can review who queried what and when.
That depends on your archive and your integrations more than on us. The API is OpenAI-compatible, so the CMS and whatever you already run keep working once you swap the key and the endpoint. Indexing the archive is the long part, and you can start with recent years and work backwards.
Yes, with the same key. There the query does go out to the model provider, so source material and anything under embargo is handled with the open models inside the perimeter. It is configured per team.
In the demo we go through your case: which models you need, where they run and what it takes to meet the regulation that applies to you.
We reply within one business day.
