Apple has announced new Mac security controls in macOS 27.3 beta 2 to give users greater oversight of applications requesting access to private files and system functions. The changes respond to growing concerns about AI agents that can operate across applications with extensive permissions.
In a developer announcement published November 16, Apple outlined changes to Full Disk Access and Device Control and Data Access permissions. The update introduces clearer warnings, more selective access options and additional visibility into how applications interact with protected information.
The company is not preventing users from installing AI agents or other software requiring elevated permissions. Instead, it wants users to understand the potential consequences before allowing an application to reach sensitive data.
Apple Mac security controls separate system permissions
One major change divides Device Control and Data Access into two settings: System & App Protection Override and Accessibility. Apple says the distinction will make it easier to understand when an application is requesting permission to bypass built-in protections.
The company is also introducing a consistent permission process. Users will select the requesting application from a list instead of navigating different approval methods.
New guidance will encourage people to grant access only to the applications or folders needed for a particular task. Users will also receive clearer explanations of the information an application could access when broad permissions are enabled.
Full Disk Access is particularly sensitive because it can expose files and other information across a Mac. Although legitimate backup and productivity software may require extensive access, the same permissions can create privacy risks when granted unnecessarily.
Apple is adding an option that lets users require approval each time an application requests certain access. This provides an alternative to granting persistent permissions, although repeated confirmation prompts could interrupt automated workflows.
A new transparency indicator will also appear when an application controls another app using System & App Protection Override. The indicator is intended to make this activity more visible while it occurs.
AI agent privacy incidents prompted closer scrutiny
The announcement follows several reports raising questions about how autonomous AI software handles personal information.
In September, technology journalist Jason Aten reported that Meta’s Muse assistant appeared to access his private messages despite his understanding that he had not granted permission.
Meta disputed that characterization, maintaining that access to Messages requires users to enable the necessary permissions. The disagreement raised questions about whether permission requests adequately communicate what an AI application can do.
Separately, Wired reported a security vulnerability in ChatGPT’s Mac application that could have exposed conversation histories and other sensitive information to attackers.
The vulnerability required malicious software to already be present on the affected computer. OpenAI acknowledged the problem and released a fix in September.
These incidents illustrate different risks. One concerns whether users understand an application’s access to personal information. The other involves attackers potentially exploiting trusted software to reach protected resources.
Developers must adapt to more selective access
Apple is advising developers to update their onboarding procedures and ensure applications keep functioning when users grant access only to selected folders or apps.
The company will also offer an entitlement process for certain software categories. Backup utilities that depend on Full Disk Access and assistive technologies requiring Accessibility permissions may qualify for exceptions.
Apple said existing permissions will remain valid after the update. Applications already known to the system can also retain their current experience without immediately obtaining additional entitlements.
The changes come as more AI tools run directly on personal computers and may need access to documents, browsers, and other applications to complete tasks.
Apple’s new controls are designed to make those permissions easier to understand and manage. They do not eliminate the security risks associated with highly privileged software, but they provide users with more opportunities to limit access before granting it.










