Safety & Privacy
ÒreAyò is being built with privacy, safety, and cultural respect as foundational requirements not features to add later.
Core principles
These principles guide every decision about how ÒreAyò is designed, built, and deployed.
Privacy by design
Cultural respect
Honest about limitations
No hidden surprises
How we think about data
ÒreAyò is in active research and development. One capability — speech recognition — has a public preview that runs during scheduled testing windows, so some data is collected when it is open. Everything else remains in development.
These principles govern how that data is handled:
- • Data is collected only when necessary for functionality, not for convenience
- • Speech data is handled with the same care as sensitive personal information
- • Users will have clear control over their data and how it's used
- • Language data from Nigerian communities is treated as culturally significant, not just training material
- • Transparency about data retention, deletion, and usage is built in by default
Safety in practice
Safety is not just about data protection. It's about building AI that people can trust to behave predictably and respectfully.
This means:
Clear communication about capabilities. The system will not imply it can do things it cannot. Limitations are stated honestly.
Designed for everyday use. ÒreAyò is being built to be used by parents, elders, and families people who should not need to understand AI to use it safely.
Tested with real-world scenarios. Evaluation includes how the system behaves in noisy environments, with mixed languages, and with diverse accents not just clean test data.
No hidden behaviour. The system should behave the way people expect it to, without surprising actions or unexpected results.
Research ethics
During the research and development phase, ethical practices are especially important.
This includes:
Responsible data sourcing
Speech data used for research is ethically sourced, properly licensed, and handled with cultural sensitivity.
Transparent progress
Research updates are shared publicly without exposing sensitive details that could compromise privacy or safety.
No exploitation
Nigerian language data is not treated as a resource to extract. Communities deserve respect and benefit from the systems built with their languages.
Long-term accountability
Decisions made during research affect future users. Ethics cannot be an afterthought.
Where we are now
The speech recognition preview on our homepage converts speech in Yorùbá, Ìgbò, and Hausa into written text. It runs during scheduled public testing windows, and the page shows whether it is currently open. It is an early preview of one capability, not the finished assistant.
Recordings are not kept. Audio is used to produce your transcript and then discarded. We keep a recording only if you explicitly ask us to, using a checkbox that is never ticked for you.
If you tell us how accurate a transcript was, we store that assessment and any correction you provide. This is how we measure whether the system works with real speech rather than only with test data.
Showing a recording publicly is a further, separate choice. It never happens because someone agreed to let us keep a recording for research, and nothing is published without a person reviewing it first.
The rest of ÒreAyò — conversation, spoken answers, voice actions — is not publicly available, and no data is being collected for it.
The privacy page sets all of this out in detail, including how to ask us to delete something.
Questions about safety or privacy?
If you have concerns or questions about how ÒreAyò handles safety, privacy, or cultural respect, we want to hear from you.
Clear communication about these topics is part of building trust. Feedback helps make the system better.
Follow the journey
Join the waitlist to receive updates as safety and privacy practices are implemented, or follow research progress.
We share openly and honestly, without hype.