Standards
How we use AI
We cover AI, and we use it. Both call for the same things: say plainly where it helps, keep a person accountable, and fix our mistakes in the open.
Published by GEN:A Labs, Ottawa · Editor: Cem (Jack) Guler
The labels you will see
Everything made with AI assistance carries one of these labels.
| Label | Where | What it means | The human role |
|---|---|---|---|
| AIAI-assisted | “From our desk” pieces (made in Buildgena Studio) | Drafted with AI assistance from the sources listed under the piece. | A named editor edits and approves each piece before it is published. The approver and the date are shown on the page. |
| AIAI-assisted summary | Newsroom summaries (/dispatch and /article pages) | Summaries of what other outlets reported, written with AI and checked automatically against every report they cite. Each one links to those reports. | This is routine, automated coverage under editorial rules set by the editor. Not every summary is reviewed by a person before it appears; sensitive topics can be held for review, and the editor answers for corrections. |
| AIAI desk headline | Some headlines on story cards and story pages | A short headline and brief written with AI from the publishers' own headlines. An automatic check rejects it if it adds a number or a name the sources don't contain. The publisher's original headline is always shown next to it. | Automated, under rules set by the editor. Report anything wrong and we fix it. |
| AIAI translation | Headlines shown in the other language | Machine translation between English and French, with the original headline available. | Automated. The link always goes to the publisher's original. |
No AI label: other publishers' headlines (shown as published, with a link), and our signed guides and editor's columns. Explainer videos say on their own page when the voice is AI-generated.
What we never do
- Invent news, quotes, numbers or sources. Every fact comes from a named source, and company statements are shown as claims.
- Make pictures or video that look like news photos or footage. Our explainers use drawn, code-made scenes, and reconstructions are labelled as reconstructions.
- Show real people's faces in AI-made material, or put words in anyone's mouth: quoted lines come only from real records.
- Hide AI use. If AI helped make something, the page says so.
- Let advertisers or sponsors use AI to shape our coverage. They never choose, change or preview it.
A person stays accountable
Cem (Jack) Guler sets the editorial direction and the rules our tools follow, and is accountable for everything we publish. “From our desk” pieces appear only after an editor approves them, and the approver is named on the page.
Our desk pieces are made in Buildgena Studio, GEN:A Labs' newsroom app. Our site only accepts a Studio piece that cites its sources, carries its AI disclosure and names the person who approved it. Buildgena Studio
Spotted a mistake? Report it: we correct in public.
Questions
- Which AI models do you use?
- Mainly Google's Gemini models, and Anthropic's Claude when we switch it on, for writing, translating and grouping stories. Narrated explainers use AI voices from ElevenLabs. Our privacy page lists what is sent to these providers.
- Who is responsible when AI gets something wrong?
- We are. Cem (Jack) Guler, founder and editor, is accountable for everything AI Broadsheet publishes, including AI-assisted work. Report an error and we correct it in public on the corrections page.
- Do you use what I type or read to train AI?
- No. We don't build reader profiles. For our own writing, only publishers' public text is sent to the model provider. Questions typed into Ask the Keeper go to an AI model to get an answer, so leave out personal details.
Editorial standardsWhat we stand forPrivacyOwnership and funding