Google Investigating Breach to Data Security Policies by Dutch Language Expert

and transcribe “a small set of queries” to better understand languages, which Monsees called a critical part of building speech technology necessary to products like Google Assistant. The company learned that one of the reviewers violated Google policy by “leaking confidential Dutch audio data,” he said, and the company is conducting a review of its safeguards “prevent misconduct like this from happening again.” Language experts review about 0.2 percent of all audio snippets, which are “are not associated with user accounts as part of the review process,” Monsees said. Reviewers are directed not to transcribe background conversations or other noises, and only to transcribe snippets that are directed to Google, he said. He added that the voice assistant only sends audio to Google after a Google Assistant-enabled device detects a user is interacting with it by saying “Hey Google” or by physically triggering the Google Assistant. Monsees acknowledged that “rarely, devices that have the Google Assistant built in may experience what we call a 'false accept,'" interpreting other words or noises to be the wake word. Protections are in place to “prevent false accepts from occurring in your home,” he said. Consumers concerned about privacy can turn off storing audio data to their Google accounts completely or choose to auto-delete data every three or 18 months, he said.