The decisive moment comes before a sensitive conversation begins: check whether the recording will reach the cloud, then choose. For a son in London preparing to discuss a family debt in Twi and English, privacy needs to be clear while his thumb is still above the microphone.
In 2019, a contractor working on Apple’s Siri grading programme told The Guardian that workers regularly heard private recordings, including confidential conversations. Siri could be activated accidentally, and a small portion of recordings was sent to contractors who assessed the assistant’s responses. At that point, the people in those recordings had already spoken. They could not return to the instant before the microphone picked them up.
Apple suspended the programme and later changed it so users had to opt in before their audio could be used for grading. The uncertainty was resolved only after recordings had reached human reviewers and the practice became public.
That sequence matters because consent has most value before disclosure. A privacy explanation discovered later may clarify what happened, but it cannot restore an unsent recording.
The private part begins before the first word
A son in London has his phone in one hand. He is ready to ask for help drafting what he needs to say about a family debt. The conversation may move naturally between English and Twi because the subject itself moves between figures, obligation, respect and the words his family actually uses.
His thumb pauses above the microphone.
The important question is simple: if he speaks now, where will the recording go?
That question should have a visible answer. Nkomo gives the user explicit cloud-consent choices: never, ask each time or this session. The decision happens before the conversation continues, while the user can still change course.
“Never” keeps the boundary firm. “Ask each time” allows a fresh decision for every exchange. “This session” grants permission for the current session without quietly turning that moment into a permanent preference.
These options reduce a complicated privacy policy to a choice someone can make at the point that matters. They also respect the fact that sensitivity changes. A question about dinner plans and a conversation about family debt may happen on the same phone, but they do not deserve an automatic assumption that the user wants the same handling for both.
Natural speech can carry sensitive detail quickly
Code-switching is ordinary communication for people who speak Twi and English together. A sentence may begin with an amount in English, turn to Twi for the difficult family point, then return to English for a date or instruction. Forcing the speaker to select one language per turn breaks the conversation into artificial pieces.
Nkomo accepts natural Twi and Ghanaian English conversation by voice or text. In voice mode, the user can hold to speak, hear the reply and interrupt naturally. That makes the exchange feel closer to how people already talk.
It also means private context can arrive quickly. Names, debts, disagreements and family expectations may appear in one breath. Clear consent cannot sit several screens away or depend on the user remembering a setting chosen months ago.
This is the direct bridge from Siri’s grading controversy to the son’s paused thumb in London. Once sensitive audio has travelled, the most useful decision point has passed. Nkomo places the cloud choice before that point, where consent can still prevent the transfer.
The same principle applies beyond money. A family name can change what someone is willing to record, as explored in The Family Name Your Thumb Stops Before, and Where Your Recording May Go. Privacy controls need to match those ordinary shifts in context.
Control continues after the conversation
Cloud consent answers one question: what may leave the device. History controls answer another: what remains available locally afterward.
Nkomo keeps conversation history on the device under the user’s control. When history is turned off, that history is purged immediately. The user can also export their data or delete their account with one tap.
Those actions are concrete. They avoid vague promises about “respecting privacy” and let the user decide whether a sensitive exchange should remain accessible. Someone may want a useful phrasing for a call with family, then decide the conversation should not stay in their history once the call is over.
The app also shows errors instead of swallowing them. That matters around privacy choices. A failed action should look like a failed action, rather than leaving the user to assume a recording was handled as requested.
Keep your thumb still until the boundary is clear
The Siri recordings reported by The Guardian show why later disclosure is a weak substitute for earlier choice. Apple changed the grading programme, but the people whose private moments had already been captured could not revisit their original decision.
Before discussing a debt, health concern, family dispute or any other sensitive subject, pause at the microphone. Check the cloud setting. Choose “never,” “ask each time” or “this session” according to the conversation in front of you.
Then speak in the Twi, English or natural mix the moment requires. The useful habit takes only a few seconds: decide where your words may go while your thumb is still above the microphone.
Comments
No comments yet.