Apple has put its clearest warning yet about AI agents in writing. In its fiscal 2026 annual report, filed with the US Securities and Exchange Commission on November 3, the company warns that AI agents can act on their own — reaching systems they were never cleared to use, running or changing code, sending messages and completing transactions.

The warning sits in Item 1A of the Form 10-K, the section where public companies list the risks that could hurt their business. Apple says agents “may behave in unintended or unpredictable ways, including by taking actions that were not directed, instructed or authorized.” The language covers agents built by Apple and agents built by anyone else.

What Apple says agents could do

The filing is unusually specific. Apple writes that AI agents and other technologies “can take actions autonomously, including interacting with internal or external systems that in some cases they may not be authorized to access, executing or altering code, sending communications, and completing transactions.”

It goes further. Agents may exceed their permissions, defy directions or resist instructions, and work around Apple's security measures or other companies' controls. They may also be manipulated by malicious inputs into behaving differently from how they were designed to behave.

Perhaps the bluntest line concerns what happens after an agent acts. Apple says some consequences, “such as deleting data or completing transactions, may be difficult or impossible to reverse.” The filing also warns that agents “may conceal actions, hide, withhold or manipulate information, or generate inaccurate or deceptive outputs.”

A risk factor that grew fast

The warning is new, and the section around it has grown quickly. Apple's fiscal 2025 annual report had no separate AI risk factor at all. Apple first added one in a quarterly filing in May 2026, running about 199 words. In the fiscal 2026 10-K, the AI risk section runs about 628 words — more than three times as long — and mentions of artificial intelligence across the risk factors jumped from 6 to 45.

Apple also flags a limit on its own control. The filing says the company's “ability to oversee and control third-party artificial intelligence agents and other technologies may be limited,” a notable admission for a company whose business is built on controlling what runs on its devices.

The filing names no incident involving an AI agent and puts no dollar figure on the risk. A risk factor describes what a company believes could go wrong, not what has already happened.

Apple is not alone in flagging it

Apple is the latest big tech company to put agent risk on the formal record. Alphabet's quarterly filing on October 29 noted that during model testing it had “sometimes observed unexpected behavior, including autonomous agents taking actions outside of their intended scope.”

Apple has already started tightening access on the product side. In early October, it told developers it would add controls to Full Disk Access on the Mac — the setting that lets an app reach nearly everything on a computer — warning that as AI agents become increasingly capable and autonomous, “the risks associated with this level of access will grow substantially.”

The 10-K landed a day after Apple reported fourth-quarter results on November 2, the first earnings call under new chief executive John Ternus. The message from the filing is plain: as agents get more power over real systems, the mistakes they make on their own get harder to take back.