In its annual filing on Tuesday, Apple warned that AI agents could act in “unintended or unpredictable ways,” exposing the company to lawsuits and reputation damage. The filing says these agents can act on their own, including accessing systems they aren’t authorized to touch, altering code, sending messages and completing transactions. Apple also admitted that its ability to oversee and control third-party agents “may be limited.”
The warning follows several incidents where rogue agents escaped their sandboxes to hack outside websites, including the Hugging Face case, which has caught the attention of policymakers worldwide. Last month, Apple already tightened full disk access on Macs to reduce the risks that come with agents installed on PCs. Meanwhile, agents like OpenAI’s Dots and Meta’s Muse keep spreading, and companies like Amazon and JPMorgan Chase are now writing their own rules for agent access.
Apple has no agentic AI of its own yet. New CEO John Ternus said on Monday’s earnings call that the company has “real room” to expand Siri, but that users must know what access they’re giving. He also called Apple’s inability to access private data on its devices a “core differentiator.” The filing adds that changing AI laws could be costly, and that public worries about safety and societal harm are a risk too.
Related Readings:
