Devices such as Google Home, Google Assistant or Alexa of Amazon have gained popularity in recent years. Their main goal is to simplify people's lives by answering questions or controlling certain features such as music playback.
The devices are powered by activation phrases. Google customers must say "Google Agreement" to activate the device and communicate with them. A customer can request the local weather of the day or traffic and the device would provide the answer to this question if the question is well understood.
Opponents of these devices have privacy issues. Some say that people install spyware at home and at work when setting up these smart, voice-activated devices.
A recent report by the Belgian TV channel VRT NWS confirms some of these concerns, at least to some extent. According to the report, everything that is said during the activation of the devices is registered and stored by the company. Even worse, from the point of view of confidentiality, some employees listen to some of the conversations to form the algorithm.
VRT NWS journalists spoke to one of Google's subcontractors who showed them the system. According to the report, thousands of people around the world are listening to audio clips from Google. in Flanders and Holland, about a dozen people.
The subcontractor has revealed that the system is used to form the algorithm so that he understands better the "subtle differences and characteristics of the Dutch language", and that it is the same for the other listening operations and languages of other regions.
Speech recognition automatically creates a transcription of the recordings and one of the main tasks of the employees who work there is to listen to the conversations and make sure that the transcription is accurate.
Google tries to anonymize the data before it is made available to employees, but the automated anonymization is never 100% accurate. Although Google deletes user names and other credentials, it does not modify the record itself. Users who reveal personal and private information when they speak to the device have recorded them. Employees see the transcript and listen to the conversation, which can lead to the discovery of identities.
The team of journalists listened to over a thousand excerpts, of which more than 150 were recorded, although the word activation was not given clearly. The devices are programmed with a certain level of tolerance to ensure the recognition of activation words, even if the sentence is not pronounced clearly. The disadvantage is that conversations can be recorded even if participants never want to communicate with the smart device.
A statement from Google about the incident confirms that the company is reviewing "about 0.2% of all registrations".
How many smart voice activation device owners know what they say is registered by the company that makes these devices? How many people know that "real" people can listen to their conversations and would it change the way they interact with these devices or even prevent them from using them?
Now, it's your turn: Do you own or use intelligent devices activated by voice?