Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)

Apple Is Building an AI Agent and Making It Harder for Outsiders to Get Access: Permission Now Takes 14 Steps, Not 4

Apple is building an AI agent that will act on the user's behalf across apps, the browser, and the system, according to Bloomberg. The release is expected at the end of next year. Meanwhile, the company has made it harder for outside agents to get access in the macOS 27.3 beta: granting them control of the computer now takes 14 steps instead of four. Developers building agents for Mac should plan a fallback route through an API or the browser in advance.

AI-processed from TG: @ai_machinelearning_big_data (Machinelearning); edited by Hamidun News
Apple Is Building an AI Agent and Making It Harder for Outsiders to Get Access: Permission Now Takes 14 Steps, Not 4
Source: TG: @ai_machinelearning_big_data (Machinelearning). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @ai_machinelearning_big_data (Machinelearning) ↗
Editor-in-Chief
Apple job listingagents switch between the browser, terminal, and code
Key metric #2CEO Tim Cook is pushing the project and pressing Craig Federighi
CompetitorsAnthropic's Duet, OpenAI's Dots, and Google's Tasks and Gemini

Imagine you've written an AI agent that opens apps on your Mac, clicks through windows, and fills out forms on its own. For it to work, the user has to grant it permission to control the computer. Until now, that took four steps. In the macOS 27.3 beta, Apple has stretched the process to 14. At the same time, Bloomberg reports, the company is building its own AI agent that will act on the user's behalf across apps, the browser, and the operating system. For anyone building agents for Mac, this is a signal: the platform's terms are changing before Apple's own product has even shipped.

What is known about Apple's agent

The future agent is meant to carry out tasks on the user's behalf: open the right programs, move between them, and finish the job. It will also be able to write code. This is not a voice assistant that answers questions. It is an executor: you assign a task, and it works through all the windows itself.

A release is expected at the end of next year, and development is at an early stage. Apple has not officially announced anything. The details come from Bloomberg's sources, so the timeline and the feature set may change.

A job listing from the company indirectly supports these plans. It describes general-purpose agents that switch between the browser, documents, the terminal, and a code editor and coordinate subagents. A subagent is a smaller helper to which the main agent hands off part of the work. For example, one collects figures in a spreadsheet, another writes a report based on them, and the main agent makes sure the results line up.

Who is pushing the project and who it will compete with

According to the source, CEO Tim Cook is pushing the project. He is demanding more decisive action from software engineering chief Craig Federighi. That suggests agents are treated inside Apple as a priority, one that leadership is willing to press teams to deliver.

There is no shortage of competitors: Duet from Anthropic, Dots from OpenAI, and Tasks and Gemini from Google. All of them aim to be the digital worker that takes over routine tasks on a computer. Apple has something the others lack: it controls the hardware, the operating system, and the app store.

14 steps instead of four

The most practical detail concerns the present. In the macOS 27.3 beta, Apple has made it harder for competing agents to get computer control. To grant it, the user now has to go through 14 steps instead of the previous four.

Officially, this can be explained by security: an agent with the right to click and type gets almost complete control over the machine. But the timing coincides with the work on Apple's own agent. The sources do not specify what lies behind the decision, and since this is a beta, the rules may change by release.

What developers and product managers should do

First, if your agent controls the Mac through system permissions, plan for the risk that connecting it will become harder. The longer the chain of confirmations, the fewer people will make it to the end of setup. Prepare a clear step-by-step guide and a fallback route: working through an API or a browser extension.

Second, watch which interfaces Apple opens to its own agent. If the system agent gets special privileges, third-party solutions will have to compete with it on unequal terms. If access is also opened to developers, a large market of skills and apps the agent can call will emerge.

Third, don't rush to conclusions. The project is at an early stage, the release date is only an estimate, and nothing is known about the architecture, privacy, or where data will be processed (on the device or in the cloud). But the direction is clear: every major platform wants the agent to live inside it, not next to it.

Story

Key Questions Answered

What is known about Apple's agent?

The future agent is meant to carry out tasks on the user's behalf: open the right programs, move between them, and finish the job. It will also be able to write code. This is not a voice assistant that answers questions. It is an executor: you assign a task, and it works through all the windows itself. A release is expected at the end of next year, and development is at an early stage. Apple has not officially announced anything.

Who is pushing the project and who it will compete with?

According to the source, CEO Tim Cook is pushing the project. He is demanding more decisive action from software engineering chief Craig Federighi. That suggests agents are treated inside Apple as a priority, one that leadership is willing to press teams to deliver. There is no shortage of competitors: Duet from Anthropic, Dots from OpenAI, and Tasks and Gemini from Google. All of them aim to be the digital worker that takes over routine tasks on a computer.

What are the key details on 14 steps instead of four?

The most practical detail concerns the present. In the macOS 27.3 beta, Apple has made it harder for competing agents to get computer control. To grant it, the user now has to go through 14 steps instead of the previous four. Officially, this can be explained by security: an agent with the right to click and type gets almost complete control over the machine. But the timing coincides with the work on Apple's own agent.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…