Apple reveals new Mac security controls as AI agents raise privacy risks
First reported by TechCrunch ·
If you use AI agents on your Mac, you will need to grant them permissions more deliberately, and they may not always be able to act autonomously.
Apple is introducing new security controls for macOS 27.3 to address privacy risks associated with AI agents. These agents, capable of performing actions across a user's operating system, files, and applications, have raised concerns due to their broad access demands. Following incidents where AI apps were reported to have accessed user data without explicit consent, Apple is revising its security settings. The changes aim to provide users with greater clarity and control over the data these agents can access. Key updates include separating 'Device Control' and 'Data Access' into distinct settings ('System & App Protection Override' and 'Accessibility'), providing more granular permission options, and displaying a transparency indicator when an app controls another. Developers are advised to adapt their onboarding processes to account for these new measures, though exceptions can be requested for specific functionalities like disk backup or assistive technologies.
The new security measures signal a broader industry trend of increased scrutiny on AI agent data access, prompted by early privacy missteps. Apple's approach, which emphasizes user understanding and granular control rather than outright bans, could set a precedent for how other operating systems and platforms manage AI integration. This move acknowledges the utility of AI agents while prioritizing user data protection, potentially influencing developer practices towards building more privacy-conscious AI applications from the outset.
This development directly impacts users by making them more aware of the potential privacy trade-offs when enabling powerful AI tools on their devices. For developers, it necessitates a shift in design philosophy, encouraging them to build AI agents that operate effectively with limited data access or that clearly justify broader permissions. The introduction of more explicit user consent prompts and transparency indicators will likely become a benchmark for AI application development across various platforms.
AI-written summary. May contain errors.