Apple Developing In-House Cross-Device Agent, Potentially Launching by Late 2027
Apple is developing an in-house Agent capable of performing tasks across apps and devices, and is considering integrating it into its operating systems and Apple Intelligence as early as late 2027. Three job postings published from November 17 to 19 cover general computer use, Agent system platforms, large-model orchestration, and next-generation workflows. macOS 27.3 will also adjust access permissions for third-party AI tools, increasing the process for enabling third-party computer-use tools in ChatGPT from four steps to 14.
Why read this
AI score 53
Job postings reveal the technical direction of Apple's in-house Agent, as well as the trade-offs between its device ecosystem and permission management.
Topics
Body · AI translation
Zhidx reported on November 20 that, according to Bloomberg, Apple is developing its own Agent and posted three related job openings between November 17 and 19, further revealing its R&D plans.
Apple hopes to use this initiative to enter the rapidly growing Agent market and compete with technology companies such as OpenAI, Anthropic, and Meta, which have already secured first-mover advantages. In U.S. after-hours trading on November 19, Apple shares were quoted at $350.16, up $0.84, or 0.24%.
▲Apple share price movement. Source: Yahoo Finance
The three newly posted positions are Senior Engineering Manager, Agentic AI; Senior Applied AI Engineer, Proactive; and Agentic Systems Engineer, Proactive Intelligence.
The Senior Engineering Manager, Agentic AI will advance the development of a general-purpose computer-use Agent. The Senior Applied AI Engineer, Proactive will build an Agent system platform capable of independently identifying optimal architectures and continuously improving itself, while also establishing evaluation benchmarks and feedback mechanisms. The Agentic Systems Engineer, Proactive Intelligence will focus on automation and large-model orchestration, developing next-generation Agent workflows for the iPhone, Mac, Apple Watch, and iPad.
In addition, Apple plans to integrate the relevant Agent technologies into its operating systems and Apple Intelligence, creating an Agent ecosystem capable of performing tasks across applications and devices. Bloomberg said Apple is considering launching the Agent product as early as the end of 2027.
While advancing the development of its own Agent, Apple is also adjusting third-party AI tools’ access permissions on the Mac.
An announcement published on Apple’s website on November 16 said that the company would modify two categories of permissions in macOS 27.3 beta 2: device control and data access, and full disk access.
▲Announcement on Apple’s website. Source: Apple
Device control and data access permissions will be divided into two options: system and application protection coverage, and accessibility. Users will be able either to restrict an application’s access to specific software or folders or to require it to request authorization again each time it is used.
Apple will also add permission-use notices to give users a clearer understanding of what data an application can access or control after receiving permission. When an application uses the relevant permissions to control another application, the system will also display a corresponding indicator.
Apple said these adjustments are intended to make permission management more transparent, help users better understand authorization risks, and allow them to decide which applications may access data on their devices.
However, the adjustments will not revoke permissions that users have already granted to applications. Apple said the new permission process in macOS 27.3 will apply to new requests approved by users afterward, while existing authorizations will remain valid.
According to a November 20 Bloomberg report, Apple has also made it somewhat more difficult for third-party Agents to control Macs: in the macOS 27.3 beta, ChatGPT users must complete 14 steps to enable third-party computer-use tools, compared with just four steps previously.
Meanwhile, other technology companies are also continuing to improve their Agent products.
On October 21, Anthropic launched the Agent product Claude Duet and further expanded the scope of user testing. On November 17, Google upgraded the Tasks feature in Gemini and added capabilities such as placing phone calls on users’ behalf. On November 19, OpenAI began testing Dots with a small number of ChatGPT Plus users and expanded its availability among Business Standard users.
Consumer interest in Agent products continues to rise. In terms of application download rankings, Meta’s Muse ranked second on Apple’s App Store chart of free applications just yesterday, behind only TikTok; OpenAI’s ChatGPT and Anthropic’s Claude also entered the top five.
▲Apple App Store ranking of free applications in the U.S. Source: AppWinners
In the capital markets, however, the share prices of the companies concerned moved differently. In U.S. after-hours trading on November 19, Meta shares fell nearly 2%, while Anthropic shares dropped more than 3%.
▲Meta share price movement. Source: Yahoo Finance
▲Anthropic share price movement. Source: Yahoo Finance
Conclusion: Apple Bets on the Agent Market, with Its Device Ecosystem Key to Competition
As AI Agents gradually become a focal point of competition among technology companies, a growing number of businesses are investing in related R&D, and Apple also hopes to secure a place in this market. However, Agent development requires sustained investment in funding and personnel, while bringing products to market also entails time pressure. Apple must not only catch up with competitors that already have products but also address the transition between technological development and commercialization.
For Apple, the real test is not merely developing an Agent capable of performing complex tasks, but also deeply integrating it with existing devices and operating systems such as the iPhone and Mac. The ability to connect task-execution processes across applications and devices while protecting user privacy and operational security will directly affect the product experience. Whether Apple can capitalize on the strengths of its hardware and software ecosystem may be the key to its entry into the Agent market.
Sources: Bloomberg, Apple, Yahoo Finance
AI Trends