Article · AI translation
IT Home reported on November 4 that Apple warned in its annual financial report on November 3 that AI agents may behave in unexpected and unpredictable ways, exposing the company to risks of litigation and reputational damage that could, in turn, affect its business.
Apple’s report said: “AI agents and other technologies can autonomously perform actions, such as interacting with internal or external systems, sometimes even accessing systems they are not authorized to access. They can also run or modify code, send messages, and complete transactions. Agents may also behave contrary to expectations or unpredictably, and may even take action on their own without being asked, instructed, or authorized to do so.”
According to IT Home, several incidents involving AI agents acting beyond their intended limits had occurred before Apple issued its warning. For example, in an incident involving Hugging Face, an AI agent broke out of its sandbox environment and infiltrated a third-party website. Such incidents have again drawn the attention of policymakers in Washington, D.C., and around the world.
Last month, Apple announced changes to Full Disk Access permissions on Macs to reduce the risks that could arise if AI agents installed on computers obtained such access. Nevertheless, Apple warned in the report: “The company may have limited ability to oversee and control third-party AI agents and other technologies.”
Apple has yet to launch an AI agent capable of autonomously completing complex tasks. During the earnings call, however, newly appointed CEO John Ternus said in response to a question about AI agents that there was considerable room to expand Siri’s capabilities. “We think AI agent capabilities are very appealing to users. We just want to make sure users understand what access they are granting, and that these capabilities are delivered with security and privacy protections in place.”
Ternus emphasized that Apple cannot access private data on users’ devices, an important distinction between Apple and other companies. “Think about how much personal data is on an iPhone—contacts, communications, health information, financial information… Siri can draw on it to help you get information in a very secure, privacy-preserving way, and no one else can access that data.”
Apple also noted that regulations governing AI and superintelligence are still evolving, leaving considerable uncertainty about the future. As countries introduce new AI laws and regulations, Apple may need to spend substantial amounts to meet regulatory requirements. “Ethical debates surrounding automation, artificial intelligence, and superintelligence, as well as negative consumer perceptions of these technologies, also pose risks to the company, including public perceptions that they present safety hazards or may cause other harms and problems for society.”