Amazon retains Alexa’s data indefinitely and distributes it to third parties

The queries and conversations recorded by Alexa, the Amazon Virtual Assistant, are retained indefinitely unless the user explicitly requests the company to delete them. In some cases, some data remains in Amazon’s servers after such a request and is even disseminated to third-party companies.

Last May, Delaware Democrat Senator Chris Coons demanded that Amazon explain how long he keeps the audio recordings and their transcripts, and gives details of their use.

The response (New Window) from Brian Huseman, Vice President of Public Policy at Amazon, is unequivocal: Amazon retains recordings and transcripts as long as users do not delete them on their own through the Internet. Alexa application, or on the Alexa Privacy Hub website (New window) .

However, even when erased by the user, transcripts may still be hosted in “other Alexa storage systems,” Huseman said. Amazon is currently trying to eliminate as many similar transcripts from these other storage systems.

Data transmitted to third-party companies

In the case where the user interacts with a ” skill ” (the equivalent of an application) belonging to a third company, the data that has been transmitted to this company may be retained by this company.

Interactions involving a monetary transaction are an example of retained data, since companies are required to keep a complete list of these transactions for tax purposes. This means that people who ordered pizza or booked a ride with Uber through Alexa can not erase all their data.

Deleting a record in which the user asks Alexa to set an alarm or call back an event will only erase the voice file of the person’s voice, not the alarm or reminder.

Machine learning at the heart of the problem

According to Brian Huseman, Amazon keeps these recordings and transcripts to improve Alexa’s accuracy.

“Alexa is designed to be smarter every day – this is accomplished through Amazon’s cloud services, including machine learning software,” says Huseman. Our speech recognition and natural language comprehension systems use machine learning to adapt to the ways of speaking and the vocabulary of the clientele.”

The letter of Brian Huseman did not reassure Senator Chris Coons, motivating him rather to deepen his investigation. “The American public deserves to understand how their data is used by technology companies, and I will continue to work with both consumers and businesses to find ways to better protect the personal information of Americans,” said Coons. in a statement sent to CNet.

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *