Apple plans tighter consent controls for macOS Full Disk Access
Apple says it will require a “very explicit” user action before apps receive Full Disk Access on macOS as more capable, autonomous AI agents raise privacy risks.
Apple said it will add controls requiring users to take “very explicit” action before apps can receive Full Disk Access on macOS. The safeguard would require an explicit consent action before software gets access to files and protected app data across the computer.
Apple tied the change to increasingly capable and autonomous AI agents, saying the risks associated with that level of access will grow substantially. The announcement follows wider scrutiny of security risks involving AI assistants on macOS.
Full Disk Access is currently a broad, per-app grant. Apple’s Mac user guide says it lets an app access all files on the computer. That includes data from Mail, Messages, Safari and Home, along with Time Machine backups and certain administrative settings for every user on the Mac.
Apple said Full Disk Access largely bypasses its usual controls for private data so backup apps can work properly. The company said some developers are using it in ways that could expose files, mail, messages and browsing history without users fully understanding the access. Apple also said that giving the permission to communication apps can compromise the privacy of other people in those conversations.
The change is a consent safeguard, not a confirmed redesign of the permission’s scope. Apple wants to ensure that people who intend to grant the access can do so only through an explicit action and understand the risks first. The company did not say users will get category-by-category choices or that Full Disk Access will become narrower than today’s all-files grant.
Apple did not specify when the controls will arrive, which macOS release will include them, what action users will need to take, or whether existing Full Disk Access grants will remain valid or require reauthorization.
More news

MIT and Sakana AI propose SIFT to prioritize coding-agent benchmark runs

Google outlines verifiable private federated learning system for Gboard

AWS details multi-turn RL workflow for SageMaker search agents
