Protecting sensitive data in AI apps

About this Video

• Video Title: Protecting sensitive data in AI apps
• Channel: Google Cloud Tech
• Speakers: Martin Ommander, Aron Eidelman
• Duration: 00:08:50

Introduction

This video discusses how to protect sensitive data within AI applications built on Google Cloud. The speakers address concerns about data privacy and demonstrate how Google Cloud's Data Loss Prevention (DLP) API can safeguard data at various stages of an AI application's workflow.

Key Takeaways

• Google Cloud does not use customer data to train its own models. Customer data remains under the customer's control.
• Three key data flows require protection in AI apps: data going to the AI model, application logging, and responses to the user.
• The Data Loss Prevention (DLP) API is a single solution for protecting all three data flows. It can redact sensitive data from text and images and allows customization of filtering rules.
• The DLP API can be used proactively to filter data before it reaches the model and reactively to filter logs and responses to the user. The approach depends on application specifics and the desired level of data protection.
• The DLP API can also inspect and classify data in various data stores, including Cloud Storage and BigQuery. This allows for both data discovery and de-identification.