The rapid rise of artificial intelligence agents (AI Agents) also brings unpredictable security risks. Apple recently issued a rare warning in its annual financial filings that AI agents may act autonomously in "unanticipated or unpredictable ways," and may even access unauthorized systems, execute or modify code, send messages, and complete transactions. If things get out of control, the company may face legal action and even damage its reputation.
CNBC reported that Apple pointed out in its submission that AI agents and other technologies have the ability to perform tasks autonomously, but these systems may take actions without instructions, guidance or authorization, and the company's ability to supervise and control third-party AI agents may also be limited.
The warning reflects the potential risks behind the rapid development of AI agents. Unlike traditional chatbots, AI agents can not only respond to user instructions, but can also operate computers autonomously, interact with internal and external systems, and even complete multiple tasks on behalf of users. Dots launched by OpenAI and Muse by Meta are examples of using advanced AI models to perform tasks autonomously.
However, the ability of AI agents to operate autonomously also raises security concerns. Recent incidents of AI programs escaping from sandbox restrictions and attacking third-party websites have once again attracted the attention of U.S. and global policymakers. Large companies such as Amazon and JPMorgan Chase have also begun to formulate relevant policies for AI agent access services to deal with possible security issues.
Apple itself has begun to reduce related risks. In October this year, the company announced adjustments to the complete disk access permission control of Mac computers to reduce the security threats that may be caused by AI agents installed on computers gaining high privileges.
It is worth noting that Apple has not yet launched an AI agent that can perform tasks autonomously for a long time. However, when new CEO John Ternus talked about AI agents on the earnings call, he said that such tools have development potential for users, and that Siri still has a lot of room for feature expansion in the future.
Turners emphasized that Apple wants to ensure that users clearly understand what access rights they grant to AI, while taking into account security and privacy. He pointed out that the iPhone stores a large amount of personal information, including contacts, communication records, health and financial information. Siri can securely access this information and help users handle their needs without allowing others to obtain relevant information.
In contrast, in order to improve execution efficiency, commercial AI agents from companies such as OpenAI and Meta often need to obtain higher access rights to users' email and other accounts, which also makes data security and privacy protection important issues.
In addition, Apple also reminded in the document that global regulatory regulations related to artificial intelligence and "super intelligence" are still changing rapidly, and the future is difficult to predict. In order to cope with new regulations, companies may have to invest a lot of money. The ethical controversies brought about by AI automation, as well as consumers' negative perceptions of safety concerns and social impacts, may also become risks that companies must face.
