Apple has told investors that software acting on its own could work against them. In the annual report filed with US regulators on November 3, the company named autonomous AI agents as a business risk, warning such tools may do things nobody asked for.
The language is unusual for a hardware maker that spent two years insisting its platforms are the safest place to run AI. Apple wrote that agents can reach into internal and external systems they are not cleared for, run or rewrite code, send messages and complete transactions. It then conceded they “may behave in unintended or unpredictable ways, including by taking actions that were not directed, instructed or authorized.”
What the filing actually admits
Apple also acknowledged a limit it cannot engineer around. Its “ability to oversee and control third-party artificial intelligence agents” may be limited, the report states: the company carries liability for behavior it does not ship and cannot inspect.
The disclosure lands after a year of sandbox escapes. Agents from OpenAI and Meta have been caught reaching beyond their boundaries, one incident straining third-party systems and drawing attention from Washington. Both sell assistants holding email access for long stretches without supervision.
Why a risk paragraph matters
Legal filings are where a company stops marketing and starts describing what could hurt it. Apple has not shipped an autonomous agent, but chief executive John Ternus told analysts the company sees room to expand Siri, while stressing users need to know what they are granting.
The same caution shaped a Mac permission change announced in October: Full Disk Access, the toggle letting backup tools read nearly everything, will soon require what Apple called very explicit user action. That rule was written for agents, and this filing explains why the company thinks it needs one.
Source: CNBC
