FRIDAY, NOVEMBER 20, 2026 · GLOBAL EDITION (EN) · EDICIÓN LATAM (ES)

Apple develops an autonomous AI agent as Mac safeguards tighten

Job listings outline software that could browse, code and operate apps, while outside tools face new permission hurdles.

Apple develops an autonomous AI agent as Mac safeguards tighten

Why it matters: An agent that can act inside apps would give Apple a new way to compete in AI while raising the stakes around access to users’ devices and data.

By the numbers:
  • 17 November 2026 — date of a senior AI engineering job listing cited in the project details.

  • Late 2027 — reported potential target for bringing the technology to market; Apple has not set an official launch date.

  • Four to 14 steps — the reported increase in steps to enable certain ChatGPT computer-control functions in the macOS 27.3 beta.

Apple is developing an autonomous artificial intelligence agent designed to navigate websites, operate applications and write or correct code, according to the capabilities described in the company’s job listings. The project is at an early stage, and Apple has not officially announced a product or launch date. A potential market arrival in late 2027 has been reported, but remains unconfirmed by the company.

Job listing points to software that can take action

A job listing published on 17 November 2026 seeks a senior engineering director specialising in autonomous AI. The role calls for leadership of a platform that can interact with computers in ways similar to a person, using browsers, software and digital tools to complete tasks.

That would move the system beyond answering questions. Apple’s stated capabilities include browsing web pages, issuing commands and writing or debugging code in response to broader user instructions. The aim is for the agent to carry out sequences of actions that currently require people to switch between tools and perform each step themselves.

Apple is also considering systems that plan multi-stage tasks and coordinate specialised agents. In one possible workflow, separate tools could gather information, process it and prepare a document. Those functions are described as potential applications, not as features Apple has announced for a product.

Device access brings security requirements

An agent operating across apps may need access to personal information, credentials and the ability to act on a user’s behalf. Apple’s plans therefore include authorisation controls, isolated execution environments and transparency tools intended to let users monitor what an agent does.

The development coincides with stricter permissions for third-party AI tools. The macOS 27.3 beta reportedly adds warnings before external agents can control a computer. Enabling certain ChatGPT computer-control functions is also reported to require 14 steps, up from four. Apple has framed the changes as safeguards, while the extra friction raises questions about how easily outside services can operate on its devices.

Siri work is related, but not the same product

Apple is separately working on deeper Siri integration with external applications, which could let users request actions such as arranging a ride or making a reservation. Those improvements would not, on their own, amount to an autonomous agent able to coordinate complex tasks across different programs and services.

The agent project sits within a broader push to strengthen Apple’s AI strategy. Chief executive John Ternus has sought to accelerate that effort, while software engineering chief Craig Federighi has been asked to take a more ambitious approach, according to the project details. Apple has not disclosed the agent’s business model, compatible devices or final capabilities; its timing will depend on development progress.