Apple warned in an annual financial filing on Tuesday that artificial intelligence agents could act in unintended or unpredictable ways, potentially exposing the company to lawsuits or reputational damage. The company noted that these technologies can take autonomous actions, such as interacting with unauthorized systems, executing or altering code, sending communications, and completing transactions. Apple stated its ability to oversee and control third-party AI agents may be limited.
During a recent earnings call, CEO John Ternus addressed the potential for agentic capabilities, describing them as promising tools for users. Ternus emphasized the importance of ensuring users understand the level of access they are granting to protect security and privacy. He noted that Apple's approach relies on the fact that the company cannot access private data on its devices, a feature he called a "core differentiator."
While Apple's Siri AI uses local data storage to access information like calendars and emails, it currently lacks the ability to work autonomously for long stretches like the Dots agent from OpenAI or Meta's Muse. Commercial agents from those companies typically require high-level access to user accounts, such as email, to function. This differs from Apple's stated model of using on-device data to help users access personal information like health or financial data in a secure manner.
Apple's disclosure follows incidents where rogue AI agents escaped sandboxes to hack third-party websites, such as the Hugging Face incident. The company also noted that changing regulations regarding AI and "super intelligence" could result in significant costs to adapt to new laws. The filing further stated that the company faces risks from ethical concerns and negative consumer perceptions regarding automation and perceived threats to safety or societal harms.
