Apple tightens macOS permissions after AI agents exposed private messages

Apple announced it’s tightening controls around a macOS setting called Full Disk Access to protect data from AI tools, according to a post from Apple. The move comes after reports that Meta’s Muse AI agent accessed a journalist’s private iMessages after being granted that deep system permission. (Jason Aten)

Full Disk Access lets apps bypass security restrictions to read and write system files. Apple says some developers are using this access in ways that could expose files, mail, messages, and browsing history without users fully understanding the risks. The company plans to require explicit user action to grant these permissions in future updates.

This change appears to follow a specific incident involving Meta’s Muse tool. Meta CTO David Singleton clarified that Muse requires both Full Disk Access and an opt-in Messages connector setting to read content. Security researcher Patrick Wardle also demonstrated a vulnerability in Muse that could allow attackers to capture dictated audio and inject malicious prompts. (The Hacker News)

Wardle also reported a flaw in OpenAI’s ChatGPT app for Mac that could have allowed unauthorized access to chat logs. These findings highlight how agentic tools, which can interact with your mic, camera, and files, expand the attack surface if granted too much trust. (Wired, CVE)

What to check this week

If your office uses Macs with AI tools, you should watch for new prompts that require explicit approval for deep system access. Check if your current AI agents have Full Disk Access enabled. Review those permissions to ensure they only have access to the necessary data.

Apple hasn’t announced when the new controls will roll out. You can’t force the update, but you can control what you give away today. Be careful with what you allow to see your files and messages. Think about that before you click ‘Allow’.

Sources: The Hacker News.