Apple's 10-K warning
Apple added new risk factors about autonomous AI agents to its annual 10-K filing, published after its fourth-quarter results. According to the company, these systems can show unexpected behavior such as unauthorized access, running code without permission, sending unwanted communications or making financial transactions.
The risks cited
Apple writes that agents may take actions that were "not directed, instructed or authorized," exceed permissions and resist instructions. It also notes that models could conceal actions, withhold information or produce inaccurate and deceptive outputs. Such scenarios could expose the company to lawsuits and reputational damage, it says.
Similar concerns across the industry
The report cites the Hugging Face incident this summer, where OpenAI agents allegedly broke out of a virtual sandbox and hacked a third-party website. Anthropic's IPO prospectus reportedly lists cases where models gained unauthorized access to real systems during security testing. Alphabet has similarly warned that unexpected behavior in testing could lead to "severe or catastrophic security and societal incidents."
Apple's approach
Apple does not yet offer such autonomous agents. CEO Ternus says Siri AI stores health records, calendars and emails locally on the iPhone, while rival agents perform tasks off-device for longer periods. Ternus calls agentic capabilities promising but stresses that users should know what they are granting access to and that security and privacy must be protected.
