Home › Ethics & Consent
The rules we will not bend
AI voice work only deserves to exist if the humans behind it said yes, knew what they were saying yes to, and got paid fairly for it. These are the commitments every project through Voices4AI is held to.
Our commitments to voice talent
- Written, informed consent for every project. The consent record names the client, the purpose, where the voice may appear, and how long the licence runs. If any of that changes, new consent is sought - it is never assumed.
- No perpetual buyouts by default. Licences have terms. Renewal is a fresh agreement and a fresh fee.
- No resale, ever. Your recordings serve the named project. They are not repackaged into other datasets or passed to third parties you never heard of.
- Rates before recording. You see the session fee and the licence scope before you commit to anything.
- Deletion routes that exist in writing. Voice recordings are personal data under UK and EU law. You can ask what we hold, and the path to deletion is documented, not theoretical.
Our commitments to AI clients
- Provenance you can evidence. Per-contributor consent documentation, ready for procurement and due-diligence review.
- EU AI Act Article 10 alignment. Data governance - relevance, representativeness, error handling and bias examination - is designed into collection, not bolted on after.
- GDPR-aware handling throughout. Voice data is collected, stored and transferred with data protection built into the workflow.
- Honest capacity. If we cannot recruit the speaker profile or hit the timeline, we say so before you have spent anything.
And one for everyone
Everything we say about the work is what it claims to be. The voices on the roster are real humans who applied and consented. No fabricated talent, no invented testimonials, no case studies that did not happen - on a site about trusting where voices come from, that would be a strange place to start cutting corners.


Terms you can read in one sitting
Talent: join on the terms above. Clients: commission data you can stand behind.
Explore the work
Speech data services in detail
Recruited against your brief, recorded to broadcast standard, delivered with the provenance trail attached.
Speech data collection
The sourcing problem nobody quotes for, and how a project runs from brief to delivered dataset.
TTS training data
Single and multi-speaker corpora recorded for consistency, licensed with an end date.
ASR training data
Spontaneous, accented and conversational speech that sounds like your users.
Multilingual speech data
Dialects and code-switched speech, with native-speaker direction and QC.
Consent-cleared voice cloning
Voice identity work scoped, consented and paid for separately from dataset sessions.
EU AI Act Article 10
Updated for Regulation (EU) 2026/1744: new dates, live Article 50 rules, what we document.