Lingo Home

Governance & privacy

How Lingo / NativeAI is run, how we treat the people who contribute their voices, and what we open to the world. This page exists so anyone doing due diligence finds real answers, not a blank.

Who we are

Lingo is built by a small consortium working on open language technology for Cameroon's oral-first languages:

  • Flagship — research, models, and the platform.
  • KIAMA — community engagement and field deployment (agriculture/extension and health pathways).
  • Sense Africa — partnership and on-the-ground reach.

The project grew out of lingo.cm, an open French-pivot machine-translation service for Cameroonian languages, and is supported by the Klaus Tschira Foundation via the Alumnode programme.

Contributor consent

People who contribute recordings do so knowingly and are rewarded for it:

  • Contributors join a campaign and consent to their recordings being used to build

open voice data for language preservation and model training.

  • Quality contributions earn points redeemable for cash, mobile money, or goods — work

is compensated, not extracted.

  • Participation is by invitation/role (speaker, verifier, reviewer, manager) and can be

declined or stopped at any time.

Data governance

  • Personal data is minimised. We store a contributor's name and a contact (phone or

email) only to administer consent, roles, and reward redemption — never published.

  • Raw audio is not released. It is processed (trimmed, normalised, quality-gated),

and only cleaned, consented clips enter the open corpus.

  • Metadata is kept; raw audio is purged shortly after preprocessing.
  • Access to personal data is restricted to project administration.

Open-release policy

Public goods stay public. Our models (fp32 originals + int8 serving) and documentation are open on Hugging Face under CC-BY-4.0; cleaned voice data is released in versioned drops as it accumulates; contributor personal data and raw audio are never released; and any underlying third-party text (e.g. copyrighted Bible translations) is shared only where licensing permits. A hosted convenience service may exist, but the underlying public goods remain open and self-hostable. Full detail in our open-release policy and the [research log](/blog).

Privacy summary

  • What we collect: account info (via sign-in), contributor contact details, the

recordings you submit, and basic usage needed to run campaigns and rewards.

  • Why: to operate the contribution → verification → reward loop and to build open

language data with consent.

  • Retention: personal data only as long as needed to administer your participation

and rewards; raw audio purged after processing.

  • Your control: you can stop contributing and request removal of your personal data.

Contact

Reach us via WhatsApp or through lingo.cm. For data or privacy requests, contact the project team through the same channels.