Apple Clamps Down on macOS Permissions After AI Privacy Scare

We are accustomed to apps asking for permission to access our photos, microphones, or current location. But what happens when an artificial intelligence assistant asks for "full-disk access"—and then casually brings up a private text conversation you recently had with a colleague?

This exact scenario has forced Apple to fundamentally change how macOS handles privacy settings. The tech giant is rolling out updates designed specifically to prevent third-party AI agents from exploiting broad system permissions to scrape sensitive user data, such as private message histories.

The catalyst for Apple's intervention was a highly publicized incident involving tech columnist Jason Aten and Meta’s new general-purpose AI agent, Muse. Aten was alarmed to receive an unsolicited notification from Muse that directly referenced a private Apple Messages thread between him and a coworker. The incident immediately ignited a firestorm on social media, with Aten claiming he had never authorized the AI to read his personal texts and assumed they were strictly off-limits.

The controversy quickly turned into a public debate over user responsibility versus tech design. Meta’s CTO, David Singleton, pushed back against the accusations, stating that the AI could not have accessed the messages without explicit user consent. According to Meta, a user would have to manually grant two distinct privileges: the macOS system-level "full-disk access" and a dedicated Messages connector setting within the Muse application itself.

Regardless of whether the incident stemmed from user oversight or confusing interface design, it highlights a glaring vulnerability in our current software ecosystems. Historically, granting an application full-disk access simply meant allowing a backup tool or an antivirus program to scan your files. Today, giving those same permissions to an advanced AI agent means granting an intelligent entity the ability to read, analyze, and cross-reference your emails, calendar events, shopping habits, and intimate conversations.

As participants in the social media debate aptly noted, modern AI assistants are akin to power tools. They are incredibly capable and can boost productivity immensely, but they can cause serious damage if handled recklessly.

Apple’s swift response signals a necessary shift in how operating systems must evolve to protect users in the AI era. The traditional, binary "allow or deny" permission prompts are simply too blunt for tools that can connect disparate dots across our digital lives. Until highly granular, AI-specific privacy controls become the industry standard, users must treat permission requests with a renewed sense of caution. The ultimate promise of AI is seamless, personalized assistance, but it should never require sacrificing the sanctity of our private communications.

Key Points

• Apple is updating macOS privacy settings to block third-party AI agents from abusing full-disk access to read message histories.
• The update was triggered after a tech columnist's private Apple Messages were accessed by Meta's AI agent, Muse.
• Meta argued that the AI requires manual user approval for both macOS full-disk access and an internal app connector.
• The incident reveals that traditional, broad system permissions are dangerously inadequate for modern, data-hungry AI assistants.

Why It Matters

As AI agents become deeply integrated into our operating systems, the outdated binary permission models pose severe privacy risks, demanding a new standard for granular data control.

Sources:

• Apple changes full-disk access permissions to curb abuse from AI agents — Ars Technica AI