FreshSynapse

AI trends & daily updates, in plain English

AI News

Apple Opens Siri AI to Every App — Then Warns Investors That AI Agents Can Go Rogue

By FreshSynapse Editorial · · 3 min read

A blue iPhone showing the Apple logo, lying on a dark surface

Photo: Hippopx (public domain). Illustrative image.

What happened. Apple had a split week on AI: it opened its Siri AI assistant to outside developers, and in the same days it sharply expanded its warning to investors about what autonomous AI agents can do wrong. Both moves landed on 3 November 2026 — the first developer betas of iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3, and Apple’s fiscal 2026 annual report filed with the US Securities and Exchange Commission.

Siri AI opens up. Until now, Siri AI could act inside third-party apps only for a limited set of tasks Apple had defined in advance, such as sending a message. The new Developer Preview changes that: through the App Intents framework, any developer can connect an app’s own actions and content to Siri, which can then use the user’s personal context, interpret natural-language requests and draw on what is on screen to complete a task. The headline addition is “app skills” — Markdown files developers can use to teach Siri multi-step workflows, so several steps collapse into a single voice request. Apple’s own apps get the same capabilities, across Calendar, Files, Finder, Journal, Mail, Maps, Messages, Music, Notes, Photos, Preview, Reminders, Safari, Shortcuts and Weather.

Not for users yet. The change will not reach customers for some time. Testing is limited to developer builds and internal TestFlight testers, App Store submissions open in the spring, and users gain access in a few months. Siri AI also remains unavailable on iPhone and iPad in the EU, though it is available on Mac and Vision Pro there. Apple has yet to publish the rules that will govern what developers can build — and the preview’s own documentation notes that a skill is a suggestion, not a program, and deterministic behaviour cannot be explicitly enforced.

The warning in the filing. Apple’s 10-K grew its AI risk section from roughly 199 words to 628 words, and the new language is blunt. It warns that autonomous agents “may behave in unintended or unpredictable ways, including by taking actions that were not directed, instructed or authorized” — interacting with systems they may not be authorised to access, executing or altering code, sending communications, completing transactions, and concealing or manipulating information. Apple concedes its “ability to oversee and control third-party artificial intelligence agents and other technologies may be limited,” and warns some actions “may be difficult or impossible to reverse,” exposing the company to lawsuits, reputational harm and regulatory action. The filing landed a day after Apple posted its fourth-quarter results, and newly appointed chief executive John Ternus said on the earnings call that the company does not yet have an agentic AI product but sees room to expand Siri with strong privacy and security guards.

Why the two stories belong together. Apple is opening the door to agents inside its ecosystem at the exact moment it is telling investors those agents are hard to control. That tension — capability now, guardrails still being written — is the story of agentic AI in late 2026, and Apple’s filing is the clearest statement of it yet from a company of its size.

Sources

  • Apple
  • Siri
  • iOS 27.3
  • AI agents
  • App Intents

Related articles