All the recordings collected by Google’s AI assistant are transcribed by the paid contractors, thus as a result they end up listening to sensitive information about users, including names, addresses, and details about their personal lives. This information has been revealed by the Belgian public broadcaster VRT NWS in a report.
Before this report, we believed that our interactions with Al assistants are all private, but according to the latest story this is not the case. This task may be done by different companies in order to improve the company’s AI systems. Another report from Bloomberg revealed the similar facts about Amazon’s Alexa, earlier this year. This report clearly explained that how all the audio clips recorded by Echo devices are sent to human contractors, who transcribe what’s being said. And this whole process is done without users’ permission.
Some reports also say that these recording may be just because of a mistake or these audio clips are often recorded entirely by accident. Sometimes, the recording of the audio starts when AI assistants like Alexa and Google Assistant hear their wake word for example “Okay Google”
The story by VRT NWS focuses on Dutch and Flemish speaking Google Assistant users. According to this story, a thousand or more recordings were reviewed by the broadcaster, however, 153 of which had been captured accidentally. A contractor also told the publication that every week he transcribes around 1,000 audio clips from Google Assistant. He also shared his experience about one of the clips he reviewed that it was a female voice which he heard and she was in distress. He also felt that “physical violence” had been involved. Contractor also added:
“And then it becomes real people you’re listening to, not just voices,”
On this whole case, tech companies are stressing that the audio clips must be transcribed by the humans and this sending of clips to humans is essential part of process for improving their speech recognition technology. But they also insist that a very small number of clips are shared in this way. Only 0.2 percent of all recordings are transcribed by humans according to a spokesperson for Google, he told it to Wired. They also revealed that these audio clips are never presented with identifying information about the user.
A technology privacy researcher at the Alan Turing Institute in London, Michael Veale, says that these problems could cause legal trouble for the company. He also revealed that this level of disclosure might not meet the standards set by the EU’s GDPR regulations.
Company is still not talking on this issue and different tech reporters are trying to reach them. Hope so we will get any updates soon about this issue.