Apple tightens macOS Full Disk Access controls citing risks from AI agents

Apple announced on October 2, 2026 that it will implement stricter controls on macOS Full Disk Access permissions, warning that increasingly capable and autonomous AI agents pose growing risks to users' files, emails, messages, and browsing history. The decision follows complaints from Inc columnist Jason Aten that Meta's Muse app allegedly accessed private messages without permission, though Meta disputed this claim and stated access requires explicit opt-in. Industry research shows enterprise AI adoption faces trust challenges, with 60% of large enterprises slowing or postponing deployments due to reputation and regulatory concerns, while only 2% have fully implemented AI systems despite Capgemini estimating AI agents could create $450 billion in value by 2028.

macOS Full Disk Access controls

• ▪Computers running macOS allow an application that possesses the Full Disk Access feature to have much broader access to user data than iPhones and iPads.
• ▪Apple's decision to implement stricter macOS controls followed complaints about Meta's Muse.
• ▪iPhones and iPads are programmed to provide a level of separation between applications as a default protection.
• ▪Apple stated that Full Disk Access might allow certain files, emails, messages, and browsing history to be accessed, as it circumvents the safeguards that typically protect app data.
• ▪Apple will introduce additional controls to ensure that users who genuinely wish to grant an app Full Disk Access can only do so with very explicit user action.
• ▪Apple stated that as AI agents become increasingly capable and autonomous, the risks associated with Full Disk Access will grow substantially.
• ▪Apple announced on October 2, 2026 that it would implement stricter controls on macOS applications that request extensive access to data on a Mac.

Meta Muse access controversy

• ▪Meta Spokesperson Andy Stone said access to Apple Messages in Muse is opt-in.
• ▪Meta disputed Jason Aten's claim about Muse reading private messages without Full Disk Access.
• ▪Inc columnist Jason Aten complained that Meta's Muse read private messages on his Mac even though he had not enabled Full Disk Access.
• ▪Meta tightened its security warning for Muse following the discovery of a flaw classed SEV-2 by a researcher that could have potentially allowed an attacker access to a user's virtual machine, emails, and files.
• ▪Meta Muse users must first enable both Full Disk Access and the Messages connector to access Apple Messages, which they can revoke at any time.

AI agent permission frameworks

• ▪The World Economic Forum's Agent Capability and Authorization Profile is designed to make AI agent access limits easier to track and enforce.
• ▪PwC argues that companies should manage AI agents similar to a digital workforce, with clear ownership, access based on specific tasks, and limits on what they can do on their own.
• ▪The OECD says agentic AI is still evolving, and more work is needed before these systems can be considered trustworthy.
• ▪The World Economic Forum recommends an Agent Capability and Authorization Profile to define what an AI agent can access and what it is allowed to do.

Enterprise AI trust barriers

• ▪FTI Consulting research shows that 60% of large enterprises have either slowed down, stopped, or postponed the deployment of AI due to issues related to reputation, regulation, and trust.
• ▪A SAS report found that 76% of participants trust generative AI compared to 66% of participants who trust agentic AI.
• ▪Companies that invest in trustworthy AI methods have seen a 15 times greater likelihood of receiving high or good ROI from their AI efforts, according to a SAS report.

AI agent adoption forecasts

• ▪Capgemini estimates that by 2028, AI agents could create up to $450 billion in value.
• ▪Only 2% of companies have implemented AI systems fully, according to Capgemini.

AI platform investment growth

• ▪Gartner forecasts that investments in AI models and platforms will increase from $39.3 billion in 2025 to $64.3 billion in 2026.
• ▪Gartner forecasts worldwide AI platforms and models market to grow 63 percent in 2026.

Debatable claims

• ▪Apple's tightening of Full Disk Access controls is justified by AI agent risks
• ▪Operating systems should restrict AI agents' access to user data by default
• ▪Agentic AI's need for broad data access creates unacceptable security risks
• ▪Explicit permission controls for AI agents will increase user adoption