Next time you use Amazon Alexa to message a friend or order a pizza, know that the record could be stored indefinitely, even if you ask to delete it.
In May, Delaware Senator Chris Coons sent Amazon CEO Jeff Bezos a letter asking why Amazon keeps transcripts of voices captured by Echo devices, citing privacy concerns over the practice. He was prompted by reports that Amazon stores the text.
“Unfortunately, recent reporting suggests that Amazon’s customers may not have as much control over their privacy as Amazon had indicated,” Coons wrote in the letter. “While I am encouraged that Amazon allows users to delete audio recordings linked to their accounts, I am very concerned by reports that suggest that text transcriptions of these audio records are preserved indefinitely on Amazon’s servers, and users are not given the option to delete these text transcripts.”
CNET first reported that Amazon’s vice president of public policy, Brian Huseman, responded to the senator on June 28, informing him that Amazon keeps the transcripts until users manually delete the information. The letter states that Amazon works “to ensure those transcripts do not remain in any of Alexa’s other storage systems.”
However, there are some Alexa-captured conversations that Amazon retains, regardless of customers’ requests to delete the recordings and transcripts, according to the letter.
As an example of records that Amazon may choose to keep despite deletion requests, Huseman mentioned instances when customers use Alexa to subscribe to Amazon’s music or delivery service, request a rideshare, order pizza, buy media, set alarms, schedule calendar events, or message friends. Huseman writes that it keeps these recordings because “customers would not want or expect deletion of the voice recording to delete the underlying data or prevent Alexa from performing the requested task.”
The letter says Amazon generally stores recordings and transcripts so users can understand what Alexa “thought it heard” and to train its machine learning systems to better understand the variations of speech “based on region, dialect, context, environment, and the individual speaker, including their age.” Such transcripts are not anonymized, according to the letter, though Huseman told Coons in his letter, “When a customer deletes a voice recording, we delete the transcripts associated with the customer’s account of both of the customer’s request and Alexa’s response.”
Amazon declined to provide a comment to Gizmodo beyond what was included in Huseman’s letter.
In his public response to the letter, Coons expressed concern that it shed light on the ways Amazon is keeping some recordings.
“Amazon’s response leaves open the possibility that transcripts of user voice interactions with Alexa are not deleted from all of Amazon’s servers, even after a user has deleted a recording of his or her voice,” Coons said. “What’s more, the extent to which this data is shared with third parties, and how those third parties use and control that information, is still unclear.”
Clarification: Updated language to more clearly differentiate between audio records and text records collected by Amazon.