Particle.news

Apple Expands Team to Build Proactive AI Agent That Can Use Apps and Code

Stricter macOS permission steps show Apple will foreground privacy as it seeks to match existing AI agents.

Overview

  • The job listing that Bloomberg flagged on Thursday asks for a senior engineering manager to lead work on a “general‑purpose computer‑use” agent that can browse the web, control apps, run commands, and write and debug code on users’ behalf.
  • Apple’s posting describes agents that plan and carry out multi‑step tasks across browsers, apps, command lines, code, and APIs and that can coordinate specialized sub‑agents to complete workflows.
  • The company says safety, sandboxing, strong permission models, and transparency are central design requirements because agents may run code and use credentials on a user’s device.
  • Apple has tightened macOS 27.3 developer‑beta controls that increase the steps required to enable third‑party ‘computer‑use’ access from four to 14, a change the company frames as protecting users while some developers view it as limiting competitors.
  • Development is described as early with reports saying Apple has considered a public introduction as soon as late 2027, and the role would work across Siri, UI frameworks, and Apple Foundation Models as the company tries to catch up with OpenAI, Anthropic, Meta, and Google.