AI may act without permission and conceal errors — Apple warns investors
In its annual report, the company described the risks of autonomous AI agents, ranging from unauthorized actions to bypassing protective mechanisms. Potential consequences include legal claims, financial and reputational damage.
The document states that AI agents are capable of autonomously interacting with internal and external systems, including without necessary access permissions, modifying software code, sending messages, and conducting transactions. To mitigate these risks, Apple changed the full disk access settings on Mac computers last month.
The company does not yet have its own agent-based AI systems. CEO John Ternus believes that Siri can be significantly expanded with such features if data security and privacy are ensured. According to him, Apple expects that new voice assistant capabilities will encourage device upgrades and help offset rising AI expenses.
Ternus noted that Siri can work with contacts, messages, medical, and financial information on the iPhone, while Apple does not gain access to this data. For comparison, commercial agents from OpenAI and Meta request broad access to user accounts, including email.
Apple identifies evolving regulations in the AI and superintelligence sector as a separate risk: adapting to new laws may require significant expenditures. The report also mentions consumer concerns due to automation, security threats, and other societal impacts.
