An AI should ask before it sends your words to the cloud, because a spoken message can contain family details, work information, money questions, and names you did not intend to share. “Never” matters because consent only has weight when refusing is a real choice.
In 2013, the U.S. National Institutes of Health faced a difficult question around the HeLa cell line. Henrietta Lacks had been treated at Johns Hopkins Hospital in Baltimore in 1951, and cells taken during her treatment became central to medical research after her death. Decades later, advances in genetic research raised new privacy concerns for her living descendants. The outcome was not treated as automatic. NIH reached an agreement with the Lacks family that created controlled access to HeLa genomic data and included family members in the review process.
The NIH announcement is worth reading because it records a simple principle with serious consequences: access to sensitive material requires a decision about who gets to see it, under what conditions, and who has a voice in that decision.
Your voice note may not be a genome. But the underlying question is familiar: who decides where your information goes after you speak?
Consent needs to come before the first sentence
A consent screen buried after setup is too late for a voice companion. By then, you may already have said the thing you wanted to keep close: a question about a relative, an address, a work problem, or a message you are trying to phrase carefully in Twi and English.
That matters especially when code-switching is natural. You might begin in English, move into Twi for the part that carries feeling or family context, then switch back without noticing. The language mix is part of the meaning. It can also make a message more personal than it first appears.
Nkomo puts the cloud choice in front of that moment. You can choose never, ask each time, or allow cloud use for the current session. The choice is plain, and it should stay plain. “Never” means you do not have to make a speech, defend your preference, or hunt through a settings maze to keep your words off the cloud.
This is why the cloud choice must come before the first sentence is more than a privacy preference. It changes the relationship. The app asks rather than assumes.
“Never” gives a boundary its meaning
People often accept defaults because they are busy. A prompt appears while you are trying to send a quick reply, and the easiest route is to tap through it. That is not the same as making an informed choice.
A useful consent control makes the available paths understandable at the moment they matter. If you want cloud use for one conversation, choose “this session.” If you want to decide based on the question, choose “ask each time.” If the answer is no, choose “never.”
The last option prevents a common trap in digital products: a setting that sounds protective but quietly leaves room for automatic sharing later. A real boundary needs a clear edge.
Nkomo also gives you control over on-device history. Turn history off and it is purged immediately. You can export your data in one tap or delete your account in one tap. These are practical actions, not vague promises about being careful with your information.
A clear answer is better than a silent assumption
Privacy controls only help when the product is honest about what happened. If something fails, you need to know. A swallowed error can leave you wondering whether your words were sent, saved, or lost.
Nkomo shows errors instead of failing silently. That is a small design choice with a large effect on trust. You can respond to what the app tells you, rather than guessing after the fact.
The same standard applies to language. A Twi-English voice companion should not pretend every phrase has been understood perfectly. It should help you speak naturally, while being clear about the actions it takes with what you say. For a closer look at why mixed-language conversation needs care beyond a language-learning exercise, read Practical Twi practice versus natural code-switching.
Control is part of a respectful conversation
The 2013 HeLa agreement did not erase the history that led to it. It showed that access to sensitive information could be handled with more care, more clarity, and a meaningful role for the people connected to it.
That is the lesson for everyday AI. The respectful moment is not a long policy document after your data has moved. It is the short question before you speak: may this leave your device?
For some conversations, the answer will be yes. For others, it will be “never.” A companion built for natural Twi and Ghanaian English should make room for both answers, then behave accordingly.
Comments
No comments yet.