Google Clarifies: Android SafetyCore’s AI Classification is User-Controlled, Not Automatic Scanning

Google Clarifies: Android SafetyCore's AI Classification is User-Controlled, Not Automatic Scanning

Android SafetyCore: Google’s New Privacy-Focused Security Feature

Google has launched SafetyCore, a new system service for Android 9+ devices, designed to enhance user protection while maintaining privacy. The service provides on-device infrastructure for secure content classification, helping users identify unwanted content without compromising their data security.

Key Features and Functions:
– Requires 2GB RAM
– Compatible with Android 9+ and Android Go devices
– Focuses on detecting malware, messaging spam, and phone scams
– Users maintain full control over the service
– Only classifies content when specifically requested by apps

Privacy Considerations:
Unlike client-side scanning (CSS), SafetyCore does not perform automatic content scanning. Instead, it operates similarly to Apple’s Communication Safety feature, using on-device machine learning models to analyze content only when requested through enabled features.

The GrapheneOS team has confirmed that SafetyCore’s primary function is to provide machine learning models for content classification, emphasizing that it cannot be used for detecting and reporting illegal content to external services.

This implementation ensures that Android users benefit from enhanced security measures while maintaining control over their privacy and personal data.

Share This Article