Apple’s Plans Beyond Siri: An AI Agent That Can Browse, Click, and Code

Apple wants an AI assistant that can take “do it for me” literally.

Apple’s Plans Beyond Siri: An AI Agent That Can Browse, Click, and Code

Apple’s Plans Beyond Siri: An AI Agent That Can Browse, Click, and Code

Apple wants an AI assistant that can take “do it for me” literally.

A new engineering job listing reveals plans for an autonomous agent capable of operating a user’s device independently—moving Apple beyond Siri’s current capabilities and into a market where OpenAI, Anthropic, and Google are advancing their own computer-use tools.

Apple has not announced a launch date, and its plans come as the company introduces tighter macOS permissions for third-party automation tools.

For IT teams, the proposed capabilities bring application access, credential security, and oversight of automated actions into focus.

Apple’s job listing describes a general-purpose AI agent

Bloomberg first reported Apple’s plans after the company posted new AI engineering roles this week. A Nov. 17 listing for a senior engineering manager in agentic AI describes a system that could work across browsers, applications, command lines, and APIs.

Advertisement

“Imagine an AI agent that can use a computer the way a capable person does,” Apple wrote in its engineering job listing.

The proposed agent would plan and execute multi-step tasks, write and debug software, and coordinate specialized subagents. Apple also wants to test whether agents can complete lengthy tasks reliably.

A developer might eventually ask it to investigate a problem across a browser, code editor, and terminal, although Apple has not demonstrated that sequence.

The engineer would collaborate with teams responsible for Siri, user-interface frameworks, and Apple Foundation Models.

How the planned agent differs from Siri AI

Siri AI already supports personal context, on-screen awareness, and actions within compatible apps. In a Nov. 3 announcement, Apple opened a developer preview allowing third-party apps to expose actions to Siri through App Intents. Examples include checking deliveries, booking reservations, and placing orders.

Apple has already expanded Siri AI’s capabilities, but the proposed computer-use agent would offer greater independence in navigating applications and completing complex tasks.

Advertisement

The proposed computer-use agent would go further in several areas.

Capability Siri AI and App Intents Planned Apple AI agent
App interactions Uses supported actions exposed by developers Intended to navigate interfaces across applications
Web browsing Retrieves web information and supports structured app actions Could navigate websites and select tools
Coding Offers writing assistance, but not general-purpose coding automation Intended to write, debug, and execute code
Multistep tasks Can combine supported actions through developer-defined integrations Designed to plan tasks and coordinate specialized subagents
Availability Existing Siri AI features and developer preview In development, with no confirmed launch date

The main difference is how independently the proposed agent could operate. Siri AI relies on supported integrations and defined actions, while Apple’s new project aims to let software navigate interfaces and select tools to complete broader tasks.

For businesses, that flexibility could complicate access controls. An assistant that can operate applications directly may require more extensive permissions than one using predefined app actions.

Advertisement

Apple has not said whether the agent will become part of Siri or operate as a separate assistant.

Apple is also changing how AI tools control Macs

Apple is tightening how applications gain control over Mac data and other software. In a Nov. 16 developer notice, the company warned that granting broad access puts “all of the data on their Mac at risk.”

macOS 27.3 beta 2 separates certain system-control and accessibility permissions, offers more narrowly scoped access, and adds an indicator when one app controls another.

Users can also choose to approve access each time, while previously granted permissions remain active.

For IT teams, the macOS 27.3 permission changes make existing app permissions worth reviewing before allowing AI agents to access files and control other applications.

The restrictions are particularly relevant as AI assistants become capable of interacting with files, running commands, and operating applications on a user’s behalf.

Bloomberg reported that enabling ChatGPT’s computer-use tools now requires 14 steps instead of four. Apple has not explained how its future agent would request comparable permissions.

Advertisement

What IT teams should check before autonomous agents arrive

Apple’s agent is still in development, but organizations can review how existing AI assistants access applications and sensitive information.

Three areas deserve attention:

  • Review application permissions. Identify which AI tools can access files, control other apps, or execute commands. Remove permissions that are no longer necessary.
  • Set limits on sensitive actions. Require human approval for high-risk activities, such as accessing credentials, changing security settings, or executing commands with elevated privileges.
  • Test macOS permission changes. Check whether macOS 27.3 affects existing automation tools and confirm that administrators can monitor and revoke access when necessary.

Credential security also needs attention. 1Password’s integration with Claude offers one approach, allowing AI agents to sign in without exposing the underlying passwords.

Even with restricted permissions, organizations need activity records and clear approval rules to understand what an agent has done and who authorized it.

Advertisement

Bloomberg reported that Apple has considered introducing its agent as early as late 2027. However, development remains in its early stages, and the company has not confirmed a launch schedule.

Until Apple demonstrates the technology and explains its security model, IT teams should treat the job listing as an indication of the company’s direction rather than a deployment roadmap.

Thinking about putting AI agents to work on your team? Explore The Neuron Academy and its How to Talk to Your AI course to build practical AI skills and learn when human oversight is still needed.

About Author

What do you feel about this?

Subscribe To InfoSec Today News

You have successfully subscribed to the newsletter

There was an error while trying to send your request. Please try again.

World Wide Crypto will use the information you provide on this form to be in touch with you and to provide updates and marketing.