Apple named the possibility of AI agents losing control as a key business risk. As it expands AI features across devices and services, it specified the risks of technologies that autonomously perform tasks in its annual report.
Apple said in its fiscal year 2026 annual report filed with the U.S. Securities and Exchange Commission (SEC) on the 3rd (local time) that AI agents may not follow human instructions or could act unexpectedly.
The risk is not limited to its own technology. It said agents from external companies could also cause problems when they access systems to change code, send messages, or process transactions. It also flagged the possibility of bypassing security measures or being exploited by attackers.
In particular, it warned that when multiple steps are performed in succession, errors can accumulate and lead to damage that is hard to undo, such as data deletion. It also cited the uncertainty of AI-related regulation, consumer pushback, and ethical controversy as business burdens.
This warning also aligns with Siri’s expanded app integration. Apple provides an “app intent” technology that lets developers connect app features and data to AI. Through this, Siri understands what is on the screen and a user’s personal context, and performs tasks in apps such as Messages, Music, and Reminders. As AI’s role broadens from generating answers to executing real tasks, the importance of managing data access and execution permissions is also growing.
Apple tightened restrictions last month on agents’ access to Mac data. Chief Executive Officer John Turners also said at a recent earnings release that agents must clearly understand the access permissions granted by users, and that agent capabilities should be implemented on the premise of security and privacy.
※ This article has been translated by AI. Share your feedbackhere.
