Apple is working on AI agent for device control, apps, browsing and coding

Sign up now: Get ST's newsletters delivered to your inbox

Apple outlined the plans in job listings added to its website this week.

Apple outlined the plans in job listings added to its website this week.

PHOTO: REUTERS

  • Apple is developing an autonomous AI agent for device control, app use, web browsing and coding. Job listings say it will work privately and handle multistep tasks.
  • The company is moving into a fast-growing market already led by OpenAI, Anthropic, Meta and Google. Apple’s App Store already shows strong demand for AI agents.
  • Apple is also adding stricter warnings in macOS 27.3 for third-party computer-use tools. It wants stronger privacy and safety, while pushing Siri towards deeper app integration.

AI generated

Apple is developing its own autonomous artificial intelligence agent that can get things done on behalf of users, looking to enter a fast-growing market already being pursued by OpenAI, Anthropic PBC, Meta Platforms and others.

The company outlined the plans in job listings added to its website this week. In one, Apple said it is seeking a senior engineering manager for a proactive, agentic AI assistant. The system will write code, browse the web, control apps and navigate devices on a user’s behalf, according to the post.

The move follows the breakout success of Meta’s Muse, which has captured the attention of consumers in recent months with its ability to make restaurant reservations, book travel and prepare news briefings. There’s also been a proliferation of tools that let AI control Mac computers and program software. 

Shares of Meta and Anthropic slid in late trading after Bloomberg News reported on Apple’s efforts. Meta fell nearly 2%, while Anthropic dropped more than 3%. 

Anthropic PBC recently widened user testing of its own agent, called Duet, while OpenAI has expanded the capabilities of Dots. And Alphabet Inc.’s Google overhauled its agent, now called Tasks, earlier this week. Like Elon Musk’s Grok Bot, Tasks lets the system make phone calls on a user’s behalf – in addition to offering standard AI agent features.

Apple’s App Store has been dominated by AI agents, with Muse long occupying the No. 2 spot among free apps, just behind TikTok, on Nov 19. OpenAI’s ChatGPT and Anthropic’s Claude were also both in the top five.

OpenAI, Anthropic and Musk’s SpaceXSI are also behind popular developer coding tools Codex, Claude Code and Cursor.

Bloomberg News reported earlier in November that Apple was developing its own AI agent offering, but the new job listings offer fresh evidence of the company’s plans. Apple has eyed introducing the service as soon as the end of next year, but development remains in the early stages. An Apple spokesperson didn’t respond to a request for comment.

“Imagine an AI agent that can use a computer the way a capable person does,” the job listing reads. “It browses the web, works across apps, runs commands, writes and debugs code, and calls services, and it chooses the right tool for each step. It does all this privately and with the polish people expect from Apple.”

The listing says the company has a team building the “foundation for that future” and that Apple is looking for the hired manager to “make it real”. John Ternus, the tech giant’s new chief executive officer, has been pushing software engineering chief Craig Federighi to move faster and take a bolder approach to AI – with an agent being key to that effort.

“People already do their most valuable work by moving fluidly between tools: a browser, a document, a terminal, an editor, a service. We’re building AI agents that can do the same,” the listing says. The posting adds that Apple is working on agents that can “plan and carry out multistep tasks” across apps, browsers, command lines, code, and application programming interfaces, and they can “coordinate specialized subagents along the way.”

The company strongly indicates its platform would be capable of autonomously navigating the interfaces of its operating systems. The listing specifically mentions development of “general-purpose computer-use agents,” the technology that allows ChatGPT and Muse to control a Mac. Google Gemini added such functionality earlier on Thursday.

Another central component is privacy, with Apple saying that safety needs to be built into the system’s design. “Agents that can run code, use credentials, and act on a user’s behalf need strong permission models, sandboxing and transparency.”

Sandboxing refers to a process that lets engineers test software without the risk of it breaking out into external networks, the rest of the operating system or the broader internet.

As the company develops its own agent, Apple is also making it more difficult for competing bots to gain broad control over its computers. In the beta version of the forthcoming macOS 27.3 operating system, the company has introduced new warning screens that prompt users to reconsider granting third-party agents permission to control their Macs.

With the change in 27.3, ChatGPT users now need to engage in 14 distinct steps to enable support for third-party computer-use tools, up from four. While Apple has positioned the shift as a move to protect customers, some AI companies and users see it as an attempt to slow down the competition. 

Apple has already been sued by the US Justice Department for actions such as blocking third-party access to features and hardware.

While Apple is still far from introducing its own full-fledged AI agent, it has taken steps to integrate its Siri digital assistant more deeply into third-party apps. The company’s upcoming iOS 27.3 release lays the groundwork for deeper collaboration between Siri AI and outside applications, allowing the assistant to, say, book a car or restaurant reservation.

Still, the updated system doesn’t go as far as a full-fledged agent, which can autonomously carry out complex tasks across multiple apps and services in the background with little additional instruction from the user. BLOOMBERG

See more on