Google announced new privacy policies around how it handles recordings made with its voice AI platform. Under the new default for Google Assistant, the company will not retain user recordings. Customers can opt into Google’s Voice & Audio Activity program when they set up the AI. This program means snippets of audio from a user may be used to help Google’s teams to improve the language understanding and processing of Google Assistant through human review. Google added that it will now delete any audio data that is recorded through an accidental activation.
This development follows a sweep of exposés and questions about how tech companies handle recordings made by their voice assistants. Reports revealed that Google, Apple, Amazon, and Microsoft all had third-party contractors who reviewed recordings without giving users a clear explanation of how or when their activity would be heard. Most of the companies announced quickly that they would make their human reviews an opt-in part of their voice AI services.
Google’s new policies not only address the human review question, but critically also acknowledge that voice assistants can be falsely triggered to begin recording. Sounds or words that may not even appear to be that similar to a voice assistant’s wake phrase can unintentionally start a recording, which led to personal and private conversations being shared to contract reviewers and tech company employees. Google’s move to provide more transparency about the existence and handling of accidental activations is the kind of action consumers will likely need to see in order to place more trust in voice AI platforms.