Alexa devices detect wake words before processing spoken requests
Alexa devices are designed to detect the acoustic pattern of a wake word, rather than transcribe everything said nearby, according to Amazon. Once they recognise the wake word, they begin actively listening to handle a request and may record or send audio to the cloud.
Amazon says Alexa saves direct conversations, and users can mute the microphone or choose whether recordings are kept for three or 18 months, until deleted, or not retained. Deletion may reduce personalisation, and recordings held by third-party devices or services may not be removed through Alexa’s settings.
- Alexa listens for its wake word.
- Users can mute the microphone or change recording retention.
- Third-party services may retain data after Alexa deletion.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Wake-word detection lets Alexa remain ready for voice commands without continuously transcribing nearby conversation, which can make hands-free use convenient while limiting what is processed. Amazon also offers microphone muting and choices about how long recordings are kept, giving users some control over privacy and personalisation.
The case against
Even if devices are designed to detect only a wake-word pattern before actively listening, they still monitor nearby sound, and an unintended wake-up could lead to private speech being recorded or sent to the cloud. The available controls may not remove copies held by third parties, and deleting recordings can reduce personalisation, so users may reasonably question whether the safeguards offer enough control.
Read the full article at the source →
Originally published by Engadget as “The reason your Amazon Alexa devices are always listening”.