Google admits its workers can listen in on AI home devices
Google has admitted that humans can access recordings made by its Google Assistant artificial intelligence system.
The tech giant on Thursday said some of its Dutch language recordings were leaked, and that the company was investigating the breach.
The recordings were obtained by the Belgian public broadcaster VRT, which reviewed more than 1,000 audio clips and found 153 had been captured accidentally.
Google Assistant begins automatically recording audio when prompted by a user, usually by saying a wake-up word or phrase like, “OK, Google”.
According to Google, contractors listen to recordings to better understand language patterns and accents, and notes that recordings may be used by the company in its user terms.
The feature can be turned off, but doing so means Assistant loses much of its personalised touch.
In June, VRT journalist Tim Verheyden contacted a couple in Waasmunster, Belgium, saying that he had gained access to the file and more than 1,000 others from a Google contractor who is part of a worldwide workforce paid to review some audio captured by the assistant from devices including smart speakers, phones, and security cameras.
Verheyden played the couple an audio file with the voices of their son and baby grandchild, as captured by Google’s virtual assistant on a smartphone.
One recording contained the couple’s address and other information suggesting they are grandparents.
A spokesperson for Google told Wired that only 0.2 per cent of all recordings are accessed by humans for transcription.
While most of the recordings found by VRT were intended, the broadcaster said around 150 of the recordings were made by mistake, after the assistant was incorrectly activated from mishearing its wake word.
However there are now concerns those recordings may not be in compliance with the EU’s General Data Protection Regulation that is intended to limit the data companies based in the EU or doing business in the EU can hold on consumers.