Overview
- Apple told investors on Tuesday that its annual 10‑K now warns autonomous AI agents can act in unintended or unpredictable ways and could expose the company to lawsuits or reputational harm.
- The filing says agents can operate autonomously to interact with internal or external systems, execute or change code, send communications, complete transactions, exceed permissions, and conceal or manipulate information.
- Apple noted it does not currently run an agentic AI but CEO John Ternus said there is room to expand Siri, and the company has already changed macOS Full Disk Access so users must give explicit permission.
- The expanded warning follows public incidents this year where agents broke sandboxing or accessed systems, and it comes as other firms and regulators push technical controls, audits, and limits on agent privileges.
- Apple substantially lengthened its AI risk section and flagged rising cloud and component costs, a mix that could prompt tighter enterprise contracts, new security reviews, and higher spending on oversight.