Apple is preparing to tighten the way macOS handles Full Disk Access, a powerful permission that can give applications access to large amounts of data stored on a Mac. The company says the move is becoming increasingly important as AI agents become more capable of acting independently on behalf of users.
The announcement comes at a time when always on AI tools are gaining access to more parts of the devices they run on. Some newer AI agents can work continuously, interact with other applications and services, and carry out tasks without requiring users to guide every individual step. That convenience also creates a bigger privacy problem when those tools can access personal information stored locally on a computer.
Apple said it will introduce additional controls in macOS so that users who genuinely want to give an application this level of access will have to take a much more explicit action.
The company made the comments in a developer post published on October 2. However, Apple has not revealed when the new controls will arrive or confirmed which upcoming macOS release will include them.
Why Apple is changing Full Disk Access
Full Disk Access is one of the more powerful privacy permissions available in macOS. It exists for legitimate reasons. Backup applications, for example, may need broad access to files and other protected areas of the operating system to perform their jobs properly.
The problem is that the same permission can potentially expose far more information than users realise.
Apple said some developers are using Full Disk Access in ways that could put users at risk. Depending on what an application does with that permission, it could gain access to files, emails, messages and browsing history.
That creates an especially serious concern for communication applications. A user may knowingly install an AI application, but the information available to that application could include conversations with friends, family members, colleagues and other people who never agreed to have their messages processed by an AI system.
Apple’s concern is that users may not fully understand the scale of access they are granting when they approve a powerful system permission.
The company specifically pointed to AI agents as a reason the issue is becoming more urgent. Unlike traditional applications, increasingly autonomous AI systems can inspect information, make decisions and take actions as part of a broader task.
That means a permission designed years ago for applications such as backup software could have very different privacy implications when granted to an AI agent.
The Meta Muse controversy
Apple’s announcement follows a dispute involving Meta’s Muse AI agent.
In September, Inc. columnist Jason Aten claimed that Muse had synced a large amount of data from his Mac Messages database despite Full Disk Access being disabled. Aten said the database had reached row 187,462 and argued that he had never given Muse permission to access the messages.
Meta rejected the claim.
Meta executive Andy Stone said Muse cannot read a user’s Messages database unless the user enables the required settings. He pointed to Full Disk Access as well as a separate Messages connector inside Muse.
David Singleton, an executive at Meta Superintelligence Labs, also said that accessing Messages requires multiple permission steps involving both the Muse application and macOS.
The dispute highlights the bigger issue facing Apple. Even when an application technically follows the permission model, users may not always understand what a particular combination of permissions means in practice.
For AI agents, that distinction matters even more. A conventional application might use access for one specific function. An AI agent could potentially use the information as part of a much broader workflow.
AI agents are raising the stakes
The timing of Apple’s announcement is significant because the AI agent market is moving quickly.
Meta’s Muse is one example of a new generation of AI tools designed to remain active and help users complete tasks. OpenAI also introduced its dots agents in September, with the ability to work toward user goals continuously and connect with thousands of applications.
At the same time, security concerns around AI applications on Mac computers have also received attention. Reports have highlighted vulnerabilities that could potentially expose sensitive information through AI software.
None of this means that AI agents are inherently unsafe. The issue is that their usefulness often depends on access to more information and more applications. The more an agent can see and control, the greater the consequences if that access is misunderstood, abused or compromised.
Apple’s proposed changes appear intended to make that tradeoff more obvious before a user grants access.
What remains unclear
Apple has not answered several important questions.
The first is timing. The company has not said which macOS version will introduce the new controls or when users can expect them.
It is also unclear what will happen to applications that already have Full Disk Access. Apple has not said whether those applications will need to ask for permission again or whether existing permissions will remain unchanged.
Backup software is another major question. Apple itself acknowledges that Full Disk Access is important for backup applications, so any new consent process will need to preserve legitimate uses without making backups unnecessarily difficult.
Apple has also not identified the developers or applications it believes are misusing Full Disk Access.
There is another question that could become increasingly important as Apple’s own AI capabilities expand. The company has not explained whether the same stricter consent requirements will apply to Apple Intelligence features that can use information from a Mac.
For users, the bigger message is simple. A permission such as Full Disk Access is not a routine setting when it is granted to an AI agent. It can determine how much of a user’s digital life an application can potentially inspect.
That is why Apple’s proposed changes could become an important part of the next stage of AI on the Mac. Making the permission harder to grant may add a little friction, but it could also make users much more aware of what they are allowing an AI system to see.
Follow TechBSB For More Updates
