Google Search Product Manager David Monsees has taken to the company’s blog to emphasize its commitment to privacy and clarify some of its policies around the use of audio data. The post was written after one of the company’s language experts leaked confidential Dutch audio data, with Monsees noting that the leak is a violation of the company’s security policy and that the incident is currently being investigated.
The expert in question was one of the many language reviewers Google hires to improve the performance of speech-based products like Google Assistant, and to ensure that the software is compatible with multiple languages. The reviewers are only given a small fraction (0.2 percent) of the audio samples Google collects, none of which are associated with a specific user account. Google also instructs its reviewers to ignore background conversation to focus only on requests explicitly directed to a Google product.
Of course, the latest breach took place despite those safeguards, so the incident is likely to raise privacy concerns about the use of biometrics at the corporate level. Facial recognition has received much of the attention in the wake of San Francisco’s public surveillance ban, but voice and speech recognition are increasingly coming under scrutiny, with Amazon now facing a lawsuit for gathering audio samples of children without the proper consent. The company recently introduced new Alexa commands that make it easier to delete those samples.
Monsees indicated that Google would be updating its security practices to prevent another incident from happening in the future, and instructed users about how they can turn off data recording features if they so desire. The breach could nevertheless take a toll on the public’s trust, and demonstrates why many privacy advocates are still concerned about the unregulated use of biometric information.
July 15, 2019 – by Eric Weiss