EN AR

Apple Builds Autonomous AI Agent to Challenge Meta, Anthropic

Nicole Jeffrey

Also in: Companies

Key Points

  1. Apple job listings show it is developing an AI agent that writes code and controls apps.
  2. Bloomberg reported the agent could launch late next year, with work still at an early stage.
  3. The move puts Apple into a crowded agent race already dominated by rivals on its own App Store.

The latest:

Apple is building an autonomous AI agent capable of completing tasks on users’ behalf, according to job listings posted on its website this week and reported by Bloomberg News. One listing seeks a senior engineering manager for a proactive assistant able to write code, browse the web, control apps and move across devices. Apple is targeting a launch by late next year at the earliest.

Details:

  • The listing: Apple’s job posting describes an agent that uses a computer the way a skilled person does — browsing, working across apps, executing commands, writing and debugging code, calling services and picking the right tool at each step, while preserving privacy. An internal team is laying the foundation for that system.
  • The capability: According to the listing, Apple wants agents that can plan and execute multi-step tasks across apps, browsers, command lines, code and APIs, and coordinate specialized sub-agents during execution. The company points explicitly to general-purpose computer-use agents able to navigate its own operating system interfaces independently.
  • The timeline: Bloomberg News reported earlier this month that Apple was developing its own agent, and the new listings add detail. Apple is aiming for a release at the end of next year at the earliest, though development remains early. A company spokesperson did not respond to a request for comment.
  • The leadership push: Bloomberg reported that new chief executive John Ternus is pressing software engineering head Craig Federighi to accelerate the pace of work and adopt a bolder approach to artificial intelligence, with the autonomous agent positioned as a central pillar of that effort.
  • Market reaction: Meta shares fell about 2% in late trading and Anthropic dropped more than 3% after Bloomberg News published its report on Apple’s agent plans.
  • The competition: Meta’s Muse agent has drawn consumer attention by booking restaurants, arranging travel and compiling news briefings. Anthropic recently widened testing of its Duet agent, OpenAI expanded the capabilities of Dots, and Google overhauled its assistant, now called Tasks, which can place phone calls on users’ behalf like Elon Musk’s Grok Bot.
  • Computer use: Google added computer-use capabilities to Gemini earlier Thursday, the same technology that lets ChatGPT and Muse control Mac devices. OpenAI, Anthropic and Musk’s SpaceXSI are behind coding tools widely used by developers, including Codex, Claude Code and Cursor.
  • App Store rankings: AI agents now dominate Apple’s own App Store. Muse held second place among free apps for an extended stretch, directly behind TikTok, based on Thursday’s ranking, while OpenAI’s ChatGPT and Anthropic’s Claude both placed in the top five.
  • Privacy and safety: Apple frames privacy as a design constraint rather than an add-on. The listing states that agents running code, using credentials and acting for users need tight permission systems, sandboxing and transparency — sandboxing meaning software is tested in an isolated environment, shielded from external networks and the wider system.

Between the lines:

Apple is entering a market where its competitors already occupy the top slots of its own store: Muse, ChatGPT and Claude all rank among the five most downloaded free apps. That inversion — rival agents as the main draw on Apple hardware — helps explain the pressure Ternus is reportedly placing on Federighi, and why a product still in early development already carries a target launch date.

What’s next

Watch whether Apple fills the senior engineering manager role, whether the company confirms the agent publicly, and whether its late-next-year target holds as Google, OpenAI and Anthropic keep expanding computer-use features.

What to read next