Apple changes macOS security as Meta Muse sparks AI privacy row

Apple is changing the way its Mac operating system handles data permissions following a wave of complaints that Meta’s new AI assistant, Muse, has been reading users’ private messages.

On 2 October, Apple announced on its developer portal that it will introduce stricter, more explicit controls for apps requesting "Full Disk Access" on macOS. Unlike iPhones and iPads, which lock apps in secure sandboxes by default, Macs have historically allowed users to grant software deep access to their storage to let tools like backup services work. Apple now warns that some developers are exploiting this setup to access highly sensitive files, browsing history, and messages without users fully realising it.

What triggered the changes

The policy shift follows immediate friction after the launch of Meta’s Muse on 8 September. Despite racking up 2.5 million downloads in its first fortnight and topping the US App Store, the agent quickly ran into trust issues.

A technology columnist writing in Inc magazine accused the AI of reading his private messages despite him explicitly choosing not to share them. When asked, Muse claimed it was simply syncing his "device notifications". Separately, YouTuber Matt Robb reported that Muse leaked his home address during a transaction on Facebook Marketplace.

The fallout was swift and divided. Amazon quickly pulled Muse from its marketplace over data-access concerns, while Shopify took the opposite route, opening its entire catalog to the tool.

How Meta has responded

Meta has strongly denied that its agent behaves covertly. Meta spokesperson Andy Stone said on X that Muse’s access to Apple's Messages is strictly opt-in. According to Stone, users must manually turn on both Full Disk Access and a specific Messages connector to let Muse read their chats, and they can revoke the permission at any time.

But Apple argues that the rise of autonomous software changes the calculation. "As AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially," the company wrote on its developer site, promising that future macOS updates will require "very explicit user action" before such permissions are cleared.

The battle for the device

The dispute points to a deeper, strategic battle between operating system makers and the builders of autonomous AI. Bank of America analyst Wamsi Mohan warned in a note to clients that Apple's biggest competitive threat is losing control of what users do immediately after picking up their devices. If users rely on third-party AI agents to run their lives, the value of the platform shifts from Apple to the agent builder.

The financial market has already begun pricing in this tension. Meta’s shares climbed nearly 30 percent in September on the back of its AI push, while Apple’s rose about 4 percent.

For now, Apple is using its gatekeeper status to push back, framing the changes as a necessary safety net. But as AI agents become the default interface for millions of users, the line between helpful automation and invasive surveillance remains dangerously thin.