Problems with Google Assistant Are Getting Serious
A story broke about Google utilizing the Assistant in units just like the Nest Hub and Google Home to “eavesdrop” on customers. VRT NWS, a Belgian news firm, reported listening to “greater than a thousand” leaked Assistant voice recordings that included “conversations that should never have been recorded, a few of which contain sensitive data.” The VRT NWS report and among the follow-on coverage by other outlets is flavored by the “evil, duplicitous big tech” narrative that’s usually adopted by the media when stories like this are reported. The Belgian data leak illustrates a serious issue that needs to be solved. but they do not have anything to do with Google eavesdropping or being duplicitous.
Google subcontracts this job to firms with expertise in local languages and dialects. An employee at one of those firms in Belgium gave VRT NWS access to the recordings. Among the many thousand, plus recordings were 153 that occurred when the Assistant activated without hearing the “OK/Hey Google” wake up command. A few of these mistaken recordings included bedroom speak, arguments, and business calls that should have been private. VRT NWS was additionally capable of discovering the identity of some of the speakers from the data present in the recordings.
The leaked voice records in Belgium are cause for concern. Google’s data security for recordings must be tightened up, and the Google Assistant’s false accept rate needs to be cut down. Media coverage that shapes the story in terms of giant tech secretly gathering user data follows a comfortable narrative that misinforms the reader, increases misunderstanding and directs attention away from the necessary problems that need to be solved.