When you import training data to Cloud AutoML, Google needs to store that content in order to train a machine learning model. This allows you to manage, annotate, modify the dataset, and to re-train the model without re-import. Data is stored securely on Google servers after encryption, until you delete the dataset or project.
When you send prediction content to Cloud AutoML, we must store that content for a short period of time in order to perform the analysis and return the results to you. The stored content is typically deleted in a few hours, although occasionally we will retain it for longer while we perform debugging and other testing. Google also temporarily logs some metadata about your requests (such as the time the request was received and the size of the request) to improve our service and combat abuse.
Regarding Speech to Text API, your data collected through data logging enjoys the same level of security as all other Google Cloud services. The Google security model is an end-to-end process, built over 15 years of experience and focused on keeping customers safe on Google applications like Gmail, Search and other Apps.
User data privacy in Google speech to text api and User data privacy in Google speech to text api have more explanation on this.
Hello Carl,
I do not have public documentation to answer that specific question but you may look here to Google’s White Paper to learn more about security at Google.
I hope this helps.