Apple Develops Neural Network Tool to Scan iPhones for Banned Content
Apple is preparing to deploy a machine learning system that scans individual iPhones for child abuse material. Critics warn the technology poses a significant risk of mission creep and threatens user privacy.
Apple is preparing to announce a new neural network-based tool that scans individual iPhones for child sexual abuse material (CSAM). According to cryptography professor Matthew Green, this system uses advanced machine learning techniques rather than traditional hash-matching to identify illegal images based on a database from the US National Centre for Missing and Exploited Children.
This approach marks a significant shift because it indiscriminately scans end-user devices locally, a practice that differs from previous methods where internet service providers simply blocked known illegal content at the network level. The exact capabilities of the neural matching function remain unclear, leaving open the question of whether it can detect entirely new images or only exact matches.
Security experts express serious concerns that this technology inherently invites mission creep and threatens Apple's privacy-focused brand image. While Apple will likely frame this as an incremental step following its scanning of unencrypted iCloud backups, both Western and authoritarian governments will undoubtedly view this invasive local scanning capability as a welcome precedent for broader surveillance.