AI disclosure
What is AI here, and what isn't.
Briefly is an AI product and hiding that would be both wrong and pointless — you can hear it in the first sentence. This page is the plain version: which parts are machine-made, which parts are decided by people, and what we will not do with either.
Made by AI
- The voice Every briefing is spoken by a licensed synthetic voice, used under terms that restrict how the model may be used, and it is never exposed to third parties.
- The writing Scripts are drafted by a language model from published reporting, then put through automated checks before anyone hears them. No story is invented; each one traces to sources you can open.
- The selection Which stories reach your briefing, and in what depth, is ranked by machine against the profile you set — your role, your industry, how deep you want to go.
- The follow-up answers Interrupt with a question and a model answers it, grounded in retrieved source passages and checked against them before you hear it.
Decided by people
- Which publishers we read at all No feed enters the pipeline without a review of its licence terms and whether it permits AI summarisation. That is an editorial decision, made by a person, recorded in a file. The resulting list is public.
- What happens when a script fails its checks A script that fails twice goes to a human review queue rather than to air, and yesterday's briefing stays up in its place.
- Corrections When a briefing gets something wrong, a person decides how it is corrected, and the correction is carried in the next briefing rather than quietly edited away.
- Refunds and support Email hello@brieflylab.ai and a person answers. Refund requests are handled by a human, not a form.
Where we tell you
The disclosure is not buried in a settings screen. It is on this website and in the app, before you have paid anything. The disclosure is shown on screen before your first briefing and before your first voice session; every briefing screen carries an AI-narrated label. If you ever cannot tell whether you are hearing a person or a model, we have failed at this — say so and we will fix it.
How to verify our audio
Every daily briefing carries C2PA Content Credentials signed by Briefly and an inaudible AudioSeal watermark, so any C2PA-aware tool can confirm the audio was generated by our pipeline and when. Live voice answers and on-demand briefings are labelled as AI-generated in the app; we are extending the same machine-readable marking to them.
What we don't do
- No synthetic voices of real people The hosts are licensed voices. We do not clone a public figure, and we do not put words in anyone's mouth.
- No advertising You pay for the briefing. Nothing in it is sold, and no story placement is for sale.
- No claim without a source A statement that cannot be traced to published reporting does not ship. This is enforced in the pipeline, not left to the model's judgement — the checks are described in full.
The honest limits
Language models get things wrong. Retrieval and citation checks reduce that substantially; they do not reduce it to zero, and a product that told you otherwise would be lying. Treat a briefing as a well-sourced starting point — that is why every story keeps a link to the reporting underneath it, and why interrupting to ask is a first-class feature rather than a novelty.
Hear it for yourself.
Invite-only for now. 3 days free once you're in, then $5.99/week. Cancel anytime.